High-throughput phenotyping of lettuce seedlings is highly prone to background confusion because the seedlings are small, have weak textural features, and exhibit spectral reflectance similar to that of the substrate. Traditional single-visual-modality approaches struggle to achieve reliable structural and physiological characterization simultaneously under the repetitive backgrounds and dense arrangements typical of greenhouse tray cultivation. To address these challenges, we establish a multimodal 3D phenotyping framework tailored for controlled agriculture environments, enabling the quantification of structural and physiological characteristics of lettuce seedlings. This framework is based on an unmanned ground vehicle (UGV) platform integrating a RGBD camera and a quad-band multispectral sensor which are rigidly coupled and synchronously triggered. An alignment module based on established feature matching algorithm is introduced to register the misalignment between source multispectral and RGBD images. Subsequently, we design a novel dual-backbone instance segmentation network, MS-SegNet, to enhance segmentation accuracy by hierarchically fusing geometric information with multispectral features. A robust 3D metric pose estimation pipeline, incorporating standard SfM initialization, scale recovery, and generalized ICP refinement, is constructed to generate 3D point clouds with spectral attributes and semantic labels. Finally, key structural and physiological phenotype parameters of each seedling are calculated based on the 3D semantic multispectral point clouds. Experiments demonstrate that MS-SegNet achieves significant advantages in instance segmentation of lettuce seedlings with mAP@50:95 = 0.854. The metric 3D pose estimation pipeline exhibits reliable performance under complex controlled conditions. The quality of the 3D reconstructions is indirectly validated through downstream structural trait extraction. The estimated seedling height and crown width show high correlation with manual measurements, achieving R 2 values of 0.8379 and 0.918, and RMSE values of 10.94 mm and 11.56 mm, respectively. Overall, by systematically integrating these adapted components with the novel segmentation architecture, this framework achieves stable performance improvements in 3D reconstruction, instance segmentation, and phenotypic analysis under greenhouse conditions. It provides a scalable, integrated technical solution for non-destructive, high-throughput phenotyping of crop seedlings in controlled environments.
Accurate and non-destructive assessment of drought stress is important for improving lettuce production and supporting timely crop management. This study presents a detection-guided deep learning framework for plant-level drought-stress assessment in hydroponically grown lettuce using bird’s-eye-view RGB images. The study further investigates whether canopy segmentation can improve classification performance by reducing irrelevant background information. The framework was evaluated using 2190 images collected across three independent cultivation cycles in which drought stress was induced by isolating the plant root zones from the nutrient solution. In the first stage, YOLO-based object detection was used to localize individual plants, with YOLO26m achieving the highest detection performance of 99.4% mAP@0.5. The detected regions were subsequently used as spatial prompts for zero-shot canopy segmentation using the Segment Anything Model (SAM), with SAM ViT-B achieving a mean IoU of 0.9864. Six convolutional, transformer-based, and hybrid classification architectures were then evaluated independently using YOLO-cropped and SAM-segmented plant images. Segmented inputs consistently improved classification performance, with MaxViT-S achieving the highest binary test accuracy of 96.3%. The framework further distinguished time-defined pre-stress, early-stress, and late-stress periods with an accuracy of 92.4%. Plant-level generalization was further assessed using six-fold leave-one-plant-out cross-validation, resulting in a mean test accuracy of 90.25 ± 1.78% on unseen plants. These findings demonstrate that RGB-based plant-level analysis can support non-destructive drought-stress assessment and that canopy segmentation improves classification by reducing background influence.
Optimizing environmental inputs for indoor crop production by conducting a traditional endpoint growth analysis requires significant time and resources. The most common scientific approach to assessing crop response involves the accumulation of dry mass at the end of a cropping cycle. A growth dynamics analysis also results in the accurate estimation of the crop response to the growth environment through periodic destructive sampling. Measuring crop gas exchange in the same environment in which it is grown offers a powerful alternative to accelerating the environmental optimization process, especially for vegetative crops. This work introduces Minitron III, a third-generation technology advancement capable of continuous gas-exchange monitoring from seed to harvest for small specialty crop stands. For proof of concept, 24 ‘Rouxai’ red oakleaf lettuce plants were grown from seed to harvest over a 25-day cropping cycle. Instantaneous differences in the carbon dioxide (CO 2 ) and water vapor (H 2 O V ) mole fraction between sample/reference lines flowing through/around cuvette/growth space were measured using a differential infrared gas analyzer, allowing determination of net photosynthesis based on a 0.41-m 2 cropping area. Crop stand net photosynthesis was detectable 7 days after sowing seeds, increasing gradually from 0.13 to 0.60 µmol·m −2 ·s −1 over the following week. The crop net photosynthesis rate increased robustly on a daily basis from 15 days after sowing seeds. While the net photosynthesis rate at the beginning of the photoperiod was 0.68 µmol·m −2 ·s −1 on day 15, it increased to 7.7 µmol·m −2 ·s −1 by day 25 after sowing seeds. Crop dark respiration was detectable from 17 days after sowing seeds and ranged from −0.3 to −0.9 µmol·m −2 ·s −1 . Minitron III has potential for rapid optimization of multiple environmental inputs for indoor production of specialty crops based on the near-real-time crop response to environmental inputs.
The increasing demand for agricultural production requires reliable and non-destructive methods for monitoring plant physiological conditions in real time, particularly in controlled environments. Spectral sensing in the visible to near-infrared (VIS–NIR) region offers a promising approach; however, the performance of low-cost sensors is often limited by calibration accuracy, wavelength-dependent sensitivity, and insufficient validation against plant physiological indicators. This study aims to evaluate the performance of a VIS–NIR plant monitoring system based on the AS7265x multispectral sensor in a controlled hydroponic environment. Lettuce (Lactuca sativa L.) was grown under two nutrient concentrations (600 ppm and 1200 ppm), and spectral reflectance data were collected across the 410–760 nm range. Sensor measurements were calibrated and validated against a LI-COR LI-180 reference spectrometer using linear regression, with performance assessed using the coefficient of determination (R²), root mean square error (RMSE), and spectral response consistency. The results show strong calibration performance, with wavelength-specific R² values ranging from 0.9682 to 0.9870 and RMSE values between 0.26965 and 5.19772. Although a systematic offset was observed, the AS7265x sensor preserved key spectral patterns, particularly in the green (510–560 nm) and red-edge (705–730 nm) regions. Differences in nutrient concentration were consistently reflected in both spectral responses and SPAD measurements, indicating sensitivity to plant physiological variations. These findings demonstrate that the AS7265x sensor provides reliable spectral information for relative plant monitoring and has strong potential as a cost-effective tool for plant quality assessment in controlled environments.
The ready-to-eat lettuce industry is rapidly expanding, increasing the need for reliable, scalable methods to assess seed germination and early growth under realistic soil conditions. This study presents an automated imaging-based approach for quantifying germination dynamics and seedling vigor using a low-cost multi-camera system under greenhouse conditions. Lettuce seeds were grown in soil either inoculated or non-inoculated with the soil-borne pathogen Rhizoctonia solani. Top-view images were acquired using commercial surveillance cameras and processed through a calibrated pipeline including geometric correction, color normalization, vegetation segmentation, clustering, and temporal tracking of emergence events. Seedling vigor was quantified through projected leaf area estimation. The proposed method enables accurate estimation of germination kinetics and growth dynamics under field-like conditions. Automated counts were validated against manual measurements at both intermediate and final time points, achieving high agreement in both cases. At the final assessment, the method reached R² = 0.98 and RMSE = 1.12, while at the midterm evaluation it achieved improved performance with R² = 0.998 and RMSE = 0.5, reflecting the lower complexity of plant structure at earlier growth stages. Results showed that pathogen inoculation significantly reduced both germination rate and seedling vigor, with up to 70% reduction in biomass accumulation. The proposed framework provides a robust, low-cost solution for high-throughput phenotyping of early plant development in soil-based systems, supporting scalable agricultural experimentation.
Sebastian Tonn · Mon-Ray Shao · Jos de Wit · Rami Mousa · Iñigo Bañales · Roy A M van Beekveld · Arjen N Bader · Henriëtte D L M van Eekelen · Ric C H de Vos · Eefjan Breukink · Guido Van den Ackerveken
Abstract Effective plant disease phenotyping is crucial for resistance breeding, but traditional visual assessment is often inaccurate and inefficient. This is particularly challenging when breeding lettuce (Lactuca sativa) for resistance to downy mildew, given the obligate biotrophic lifestyle of the causal pathogen Bremia lactucae. We discovered that B. lactucae-infected lettuce exhibits patches of increased blue-green fluorescence (BGF) under UV-A excitation from 6 d post-inoculation, preceding visible symptoms. Co-localization of BGF with hyphae, visualized with trypan blue, indicates that BGF is induced by downy mildew colonization. We therefore investigated its potential for non-invasive disease detection and quantification, as well as the underlying physiological changes. Using a custom imaging system, we demonstrate that BGF leaf area correlates with downy mildew severity and can be automatically quantified via a U-Net-based convolutional neural network, enabling early, objective disease assessment. Exploring transcriptomic and metabolomic changes associated with BGF, we found that induction of the phenylpropanoid pathway led to accumulation of caffeoylquinic acids, whose fluorescence spectra overlap with that of BGF tissue, supporting the hypothesis that these compounds contribute to the fluorescence signal. BGF imaging offers a powerful tool for phenotyping in lettuce breeding and for identifying quantitative resistance traits that support durable downy mildew resistance.
Accurate high-throughput evaluation of seed germination under abiotic stress is often hindered by subjective manual scoring and insufficient temporal resolution. This study introduces an integrated phenomics framework leveraging an explainable deep learning model for real-time monitoring of lettuce (Lactuca sativa L.) germination dynamics under salinity stress and nano-silicon priming. Utilizing a custom X-Y motorized imaging system, we captured continuous time-lapse data across 16 treatment combinations (0-60 mM NaCl × 0-300 mg L⁻¹ nano-SiO₂). We developed an ultra-lightweight architecture, YOLO26n-Ghost-EMA, which integrates Ghost convolutions and Efficient Multi-scale Attention. This model achieved 99.46% mAP@50 with a 4.5 ms inference time, providing a high detection accuracy while maintaining a lightweight architecture and favorable accuracy-efficiency trade-off compared with standard YOLO variants. while reducing computational demand by 35-50%. To ensure biological validity, Explainable AI (XAI) via Grad-CAM confirmed that the model precisely targets radicle protrusion zones, eliminating 'black-box' opacity. Response Surface Methodology (RSM) quantified the potent ameliorative effect of nano-SiO₂, identifying 100 mg L⁻¹ as the optimal concentration to recover germination from 58.57% to 84.28% under severe salinity (60 mM NaCl). By bridging real-time computer vision and plant stress physiology, this framework provides a scalable, high-resolution solution for precision seed biology and rapid assessment of abiotic stress.
Abstract Understanding the link between genetic variation and observable traits is key to crop breeding. Hyperspectral imaging captures physiological and biochemical profiles, but current supervised methods require costly trait annotations and treat each observation as a static snapshot, ignoring the temporal dynamics of plant development. We introduce SST-MAE, a self-supervised framework that learns genotype-discriminative representations from plant hyperspectral developmental trajectories, without requiring phenotypic labels. The model learns to reconstruct masked information, capturing multiple growth trajectories. Validated on 194 field-grown lettuce genotypes across eight time points, the frozen encoder serves as a feature extractor for downstream genotype classification. SST-MAE outperforms raw spectral and linear baselines, achieving AUROC > 0.89 for anthocyanin pigmentation SNPs and 0.77 for leaf serration. The learned features are highly label-efficient, attaining near-full performance with only 30–50% of labeled data, offering a scalable pathway toward high-throughput genetic screening from image-based phenotypes.
Introduction Early-stage detection and classification of lettuce heat responses are essential for non-destructive phenotyping, yet conventional assessment mainly relies on visible symptoms and manual observation. Methods This study constructed a lettuce hyperspectral dataset comprising heat-sensitive and heat-tolerant varieties under control and high-temperature treatments, and proposed the Dynamic Selective Peak Transformer (DSPformer). DSPformer integrates edge-enhanced feature extraction, dynamic multi-scale spatial-spectral representation, Peak-k selective attention, and a confusion-aware dynamic focal loss to enhance discriminative features while reducing spectral redundancy, class imbalance, and inter-class confusion. Results Under the patch-level evaluation protocol, DSPformer achieved 96.22% accuracy, 95.55% recall, 96.35% precision, and 95.95% F1-score, outperforming the compared CNN- and Transformer-based models. Day-wise evaluation showed that DSPformer reached 82.61% accuracy on Day 1 and 96.55% on Day 3, before visible heat-stress symptoms appeared on Day 6. Under a plant-level partition protocol, DSPformer maintained robust performance with 93.76 +/- 0.49% accuracy. Additional evaluation on the Indian Pines benchmark further demonstrated the applicability of DSPformer to general hyperspectral image classification. Discussion These findings suggest that hyperspectral imaging can capture heat-stress-sensitive information beyond visual phenotypes, and that DSPformer provides a promising framework for early, non-destructive lettuce heat-response screening and hyperspectral phenotyping-assisted breeding.
Plant-driven lighting control has been proposed as a strategy to regulate supplemental light-emitting diode (LED) intensity according to real-time plant physiological status. This study developed a multiple linear regression (MLR) model to predict quantum yield of photosystem II (Φ PSII ) from environmental variables and evaluated its integration into a chlorophyll fluorescence-based biofeedback light control. The model incorporated light intensity, CO 2 concentration, air temperature, vapor pressure deficit, short-term light history, and diurnal effects. In a greenhouse validation experiment, supplemental lighting was regulated using either direct chlorophyll fluorometer measurements of Φ PSII (sensor-based control) or Φ PSII values predicted by the machine learning model (ML-based control), and compared with a constant photosynthetic photon flux density (PPFD) treatment. Both sensor- and ML-based control stabilized photochemical activity across the photoperiod relative to constant PPFD. Although plant growth did not differ among treatments, sensor-based ETR control achieved the highest energy use efficiency for LED lighting in this study. These findings demonstrate the feasibility of integrating predictive ML models into plant-based lighting control systems and indicate that sensor-based biofeedback control improved the energy-use efficiency of greenhouse supplemental lighting without compromising crop growth.
ABSTRACT While whole-genome sequencing captures millions of single nucleotide polymorphisms (SNPs) and hyperspectral imaging (HSI) enables non-destructive plant phenotyping, integrating these modalities to link genotype to phenotype remains challenging due to their high dimensionality and non-linearity. This study presents DeepPheno a deep learning framework that predicts SNP genotypes from HSI data, using model predictability as a proxy for genotype-phenotype association. HSI data were acquired from 194 lettuce genotypes under field conditions. HSI data patches (20×20 pixels × 224 spectral bands) were used to train a hybrid CNN to predict the variant of a specific SNP. The framework was validated on SNPs with known phenotypic effects (anthocyanin, leaf serration, pale pigmentation), achieving high predictive performance (AUC ranging from 0.806 to 0.935), whereas models trained on randomly shuffled labels performed at chance (mean AUC ≈ 0.51). Extending the workflow to 50 randomly selected putatively neutral SNPs, most yielded low predictability, but two showed high performance (AUC > 0.76), suggesting uncharacterized genotype-phenotype links. Explainable AI, including SHAP and Grad-CAM, identified relevant spectral and spatial features driving these predictions, particularly the green and red-edge wavelengths associated with pigment dynamics and leaf structure. These results establish a framework for understanding complex genotype-phenotype interactions in plants and extracting these links from HSI data without predefining the exact trait values. It provides an avenue for high-throughput trait discovery and description and extends the integration of image-based phenomics with plant genetics.
Reproduction assets foundThe paper's Data Availability statement deposits authors' code, scripts, and supplementary material in a public GitHub repository, including a downscaled de-identified sample dataset demonstrating the pipeline. The raw HSI/genotype datasets are proprietary under NDA and not public.Code · publicThe code, scripts, and supplementary material supporting the findings of this study have been deposited in the GitHub repository at https://github.com/frankgyan/Utrecht-University--HSI .Open asset ↗frankgyan/Utrecht-University--HSIlines:195-223Plant phenotyping relevance matchOpenAlex · checked 15 Sept 2026
Early, precise, and non-destructive stress detection is essential for maintaining crop productivity, particularly in high-density plant growth systems like controlled environment agriculture (CEA), where manual monitoring is often impractical. Using plant motion as a proxy for growth and plant health, we demonstrate a method for early, non-invasive stress detection through quantitative leaf-movement analysis in lettuce and five other CEA relevant crops. Leaf-movement dynamics under stress were imaged with a low-cost, scalable Raspberry Pi imaging setup and quantified using a repurposed open-source motion estimation algorithm; Tracking Rhythms in Plants (TRiP). Our system detected stress-induced changes in leaf-movement within 1 hour of stress, with the timing dependent on the nature of the stress. Sustained reductions in leaf-movement coincide with decreased biomass accumulation. This approach offers a non-invasive, rapid, scalable, and cost-effective solution for continuous crop monitoring, with potential for application in both terrestrial and space farming CEA systems. Abstract Figure Graphical abstract: Quantification of leaf-movement dynamics as a high-throughput proxy for plant physiological status, enabling early stress detection and timely intervention to mitigate yield penalties in CEA settings (image made with biorender.org).
Determining elemental concentrations in plant tissues is essential for physiological studies on abiotic stress. However, high-throughput routine analysis of light elements (sodium to calcium) in plants is challenging due to the need for complete sample dissolution and expensive and time-consuming inductively coupled plasma-mass-spectrometry (ICP-MS). Ion chromatography and ion-selective electrodes are low-cost methods but suffer from major drawbacks, including limited throughput and time-consuming sample preparation. This study reports on a new methodology for quantitative analysis of light elements in plants using monochromatic X-ray fluorescence (MXRF) analysis. We quantitatively assessed sodium and potassium uptake in Arabidopsis thaliana, Oryza sativa and Lactuca sativa in salinity treatments. The new method provides reliable results from samples as small as 1 mg, making it suitable for analysis at the seedling stage. This is enabled by the high sensitivity of the system and optimized sample preparation that ensures sufficient signal even at low sample masses. We tested the accuracy and precision of the technique for other light elements to demonstrate its broad applicability. The results show that the method delivers rapid, non-destructive, and extraction-free light element analysis on small samples highly correlating with ICP-MS. The monochromatic XRF method provides accurate measurements and reproducible results for studying salinity tolerance ideally suited for investigating elemental composition of early plant developmental stages, offering new possibilities for research into early stimuli responses.
Real-time monitoring of H 2 O 2 in plant tissues is useful for evaluating oxidative changes during postharvest storage, but direct on-site detection in vegetables remains difficult because most assays still require tissue disruption and laboratory instruments. In this study, a dual-signal microneedle biosensor was developed by integrating polydopamine-coated Fe/Zr-MOF nanozyme (PDA@Fe/Zr-MOF) into a gelatin/sodium alginate microneedle patch for H 2 O 2 detection in lettuce. The polydopamine coating improved the peroxidase-like response of Fe/Zr-MOF through •OH generation and also contributed to photothermal conversion under 808 nm near-infrared (NIR) irradiation. After contact with lettuce leaves, the microneedles extracted interstitial fluid and allowed H 2 O 2 -triggered TMB oxidation to be read by both colorimetric imaging and thermal imaging. The two outputs were not independent recognition mechanisms, but they provided mutually supportive information and helped reduce the influence of sample color and environmental fluctuations. The sensor achieved detection limits of 0.42 μM for the colorimetric mode and 0.34 μM for the photothermal mode. During 15 days of storage at 4°C, the sensor tracked H 2 O 2 accumulation in lettuce and showed a clear relationship with spoilage progression. These results indicate that PDA@Fe/Zr-MOF-based microneedle sensing is a feasible approach for monitoring oxidative freshness changes in postharvest vegetables.
Lettuce (Lactuca sativa L.) is an important leafy vegetable with substantial diversity in leaf topology, geometry, color, and texture, which poses challenges for germplasm identification and new variety protection. However, current approaches to complex phenotypic analysis are often limited in their ability to explicitly represent and exploit semantic relationships among phenotypic traits. To address this limitation, a knowledge graph-enhanced graph learning framework for lettuce phenotypic traits was developed. Phenotypic traits were first extracted from leaf images of five lettuce types. Based on the trait description standards of the International Union for the Protection of New Varieties of Plants (UPOV), a Lettuce Leaf Phenotypic Trait Knowledge Graph (LLPT-KG) was constructed to represent semantic associations among traits. On this basis, a Dual-Channel Relational Graph Convolutional Network (DCR-GCN) was developed to jointly integrate node attribute features and graph structural information for lettuce type classification. To improve interpretability, node- and edge-level importance analyses were further performed to identify the phenotypic traits and semantic relations most relevant to type discrimination. The proposed framework achieved an accuracy of 0.94 and a Macro-F1 score of 0.94. Compared with the best-performing single-channel graph baseline, R-GCN (Relational Graph Convolutional Network), DCR-GCN improved accuracy by approximately 9% points and Macro-F1 by 10% points. These results demonstrate that combining knowledge graphs with graph neural networks can effectively capture complex phenotypic relationships in lettuce and improve classification performance. The proposed framework provides methodological support for precise lettuce germplasm identification, digital phenotypic evaluation for new variety protection, and digital-assisted pre-screening prior to field-based DUS (Distinctness, Uniformity, and Stability) testing.
This study developed an integrated diagnostic system for tebuthiuron-induced soil ecotoxicity based on morphophysiological indicators of Mucuna pruriens, using the germination index (GI) of Lactuca sativa as a sensitive ecotoxicological validation endpoint. The experiment was conducted under greenhouse conditions using a completely randomized design with 12 treatments and 360 individual pots (independent samples evaluated via destructive sampling), which were distributed across five evaluation periods at 14, 28, 42, 56, and 70 days after sowing. Morphophysiological variables, including plant height, root length, shoot and root dry mass, chlorophyll content, nodule number, and visual phytotoxicity, were quantified and integrated with multivariate and probabilistic modeling approaches. Given the multifactorial nature of the germination index, Principal Component Analysis (PCA) was applied to identify ecological and physiological gradients associated with plant vigor, stress, and symbiotic functioning. The PCA outputs were subsequently used as inputs for Probabilistic Neural Networks (PNNs), enabling the classification and prediction of bioindicator-based ecotoxicological levels using mathematically defined low, medium, and high GI classes. Model performance was internally assessed using training and validation datasets, confusion matrices, overall accuracy, sensitivity, specificity, and ROC curves. Because no independent external dataset was available, the predictive performance should be interpreted as evidence of internal consistency rather than definitive generalizability across different soils, climates, herbicide doses, or field conditions. Multivariate analyses revealed that ecotoxicological attenuation trajectories in tebuthiuron-contaminated soils are inherently nonlinear, being structured by coordinated shifts in morphophysiological traits rather than isolated responses of individual variables. The integrated PCA-PNN framework demonstrated that aboveground traits. Particularly plant height, chlorophyll content, and shoot dry mass, were more sensitive indicators of tebuthiuron-induced stress than root traits alone. Higher GI values were associated with PCA regions characterized by increased shoot biomass, greater plant height, reduced phytotoxicity, and improved physiological performance, whereas lower GI classes corresponded to suppressed growth and multidimensional stress signatures. The progressive convergence between plant vigor and GI across evaluation periods suggests a gradual mitigation of ecotoxicological stress signals on the indicator plants, indicating transitions from acute injury to physiological adaptation states. These findings confirm that M. pruriens functions as an effective bioindicator for diagnosing soil ecotoxicological status and monitoring tebuthiuron-induced impacts. However, as tebuthiuron residues were not chemically quantified, these responses should not be interpreted as direct evidence of herbicide degradation, dissipation, or removal. These findings confirm that M. pruriens functions as an effective bioindicator for diagnosing soil ecotoxicological status and monitoring tebuthiuron-induced impacts. However, as tebuthiuron residues were not chemically quantified, the observed improvements should be interpreted as evidence of physiological adaptation and/or ecological attenuation rather than definitive proof of herbicide degradation or removal. Overall, this approach provides a robust framework for early detection of soil contamination and supports its application in monitoring and guiding soil rehabilitation processes, with potential for future validation under field conditions.
Alternatives to soil-based horticulture, such as hydroponics, have been developed to respond to food distribution concerns for dense urban centers. A new system was developed to track an individual lettuce plant's growth in a hydroponic environment, utilizing streams of measured information and available models to continuously update the growth trajectory estimates for a plant. These "digital twin" models were integrated into an operating hydroponic greenhouse, with custom horticultural and sensor hardware to grow and measure relevant information. To aid in updating model parameters, plant yield was continuously measured with a custom neural network, using RGB-D images of the plants as an input. The network, trained on a collected dataset of 1300 images, was able to estimate mass within 1.5 g of the ground-truth value. After integration into the custom system, digital twin growth projections could approximate future yield between one and four days in the future, maintaining around a 2 g forecasting error.
The increasing global food insecurity driven by climate-induced natural hazards and soil degradation has made the resilience of alternative agricultural systems a critical focus in risk management. This study presents a geospatially integrated monitoring framework, the Optimized Multi-Scale Adaptive Graph Neural Network (OMSA-GNN), designed to mitigate risks associated with nutrient instability in hydroponic and aeroponic environments. The proposed system leverages a Raspberry Pi-based IoT network to monitor complex interactions among microclimatic variables, plant physiological health, and nutrient concentrations, treating them as localized geospatial data points. To enhance decision-making under environmental uncertainty, an Improved Sparrow Search Algorithm (ISSA) is employed to optimize the predictive performance of the GNN. The OMSA-GNN model incorporates visual plant indices as a proximal remote sensing approach to enable early detection of physiological stress that may lead to crop failure. Evaluated using a lettuce growth dataset, the framework demonstrates superior performance in forecasting growth trajectories and managing resource-related risks compared to conventional static models. The results highlight a scalable approach for improving the reliability of urban food systems, where traditional land-based agriculture is increasingly vulnerable to natural hazards.
Reproduction assets foundThe paper's Data Availability statement points to a public Kaggle lettuce growth dataset used for evaluation, matching an allowed URL. No author code or model checkpoints are disclosed.Dataset · publicThe datasets used and/or analyzed during the current study are available in the Kaggle repository, https://www.kaggle.com/datasets/jurijsruko/lettuce/data.Open asset ↗Kaggle · jurijsruko/lettucehtml-lines:469-500Plant phenotyping relevance matchEurope PMC · checked 5 Sept 2026
In NASA's controlled-environment plant growth systems, early and autonomous detection of crop stress is critical for sustaining food production during long-duration space missions. Hyperspectral imaging (HSI) has proven effective for early stress detection, yet the molecular processes underlying diagnostically informative spectral signals remain poorly defined. Here, we present a two-stage phenomics-to-molecular framework to evaluate whether hyperspectral signatures associated with early drought detection correspond to coordinated molecular stress responses in lettuce. In the first stage, reflectance and fluorescence HSI were used to identify early drought detection windows in 'Dragoon' lettuce subjected to controlled water limitation over a 15-day treatment period with daily imaging. Classification models integrating reflectance and fluorescence outperformed single-modality models and achieved high accuracy as early as day after treatment (DAT) 4, reaching up to 97% at DAT 5. Partial least squares discriminant analysis (PLS-DA) identified predictive wavelengths concentrated in blue-green, red, and red-edge regions associated with chlorophyll absorption and photosystem II activity. In the second stage, independent transcriptomic and untargeted metabolomic profiles were integrated with hyperspectral signatures using MOFA2 to establish biological context. This analysis revealed a dominant drought axis characterized by early activation of ABA signaling, osmotic adjustment, phenylpropanoid metabolism, and lipid and membrane remodeling, with maximal molecular divergence at DAT 5, coinciding with peak hyperspectral classification performance. Notably, wavelengths optimized for early stress discrimination were systematically shifted toward shorter, optically efficient regions relative to those most strongly associated with downstream metabolic abundance, indicating that HSI primarily captures early structural and energetic consequences of molecular stress responses rather than direct biochemical composition. Together, these results demonstrate that hyperspectral imaging can function as a non-destructive, biologically interpretable molecular proxy for drought stress, providing a foundation for compact, hands-free sensing systems capable of distinguishing stress-specific plant states in space agriculture.
Understanding plant growth dynamics requires imaging across day-and-night cycles to quantify growth, movement and development in the aerial plant body and to capture the rhythmic nature of these processes. This requires imaging in light during the day and in darkness at night without perturbing plant physiology. Nighttime imaging has typically depended on infrared (IR) illumination, producing monochrome datasets that require specialised hardware and separate analysis pipelines when combined with daytime RGB imaging. Here, we evaluated very low-intensity green (dimG) illumination from standard LEDs as a practical alternative for colour-consistent nighttime imaging and assessed its physiological impact in Arabidopsis thaliana and Lactuca sativa (lettuce). We show that high resolution colour images can be obtained under dimG using low- cost cameras, with sufficient consistency between full-spectrum and dimG images to allow direct comparison and unified image analysis. We show that very low-fluence green light (<0.5 μmol m -2 s -1 ) does not sustain circadian oscillations of gene activity under continuous exposure and does not perturb rhythms when applied during the dark phase of diel cycles. DimG imaging enabled accurate detection of diel leaf movement profiles in Arabidopsis circadian mutants, revealing genotype-specific phase differences under varying photoperiods. In lettuce, dimG pulses and continuous dimG enabled accurate quantification of diel leaf movement without affecting growth, stomatal opening, electron transport rate or chlorophyll content. Motion profiles under continuous dimG mirrored those under darkness. Our findings establish dim green illumination as a cost-effective solution for night-time imaging, simplifying phenotyping workflows with minimal impact on physiology.
Abiotic stresses, particularly acid and salt stress, severely limit plant productivity. Conventional detection is often hindered by physiological lags and phenotypic latency. Here, we develop a machine learning-enabled implantable plant biomarker sensor (MLIPBS) for early stress diagnosis. Featuring a foldable design, MLIPBS enables conformal integration into plant tissues for continuous monitoring of H 2 O 2 , K + , and pH. We confirm the robust sensing capabilities and favorable biocompatibility of MLIPBS through cross-species validation in lettuce, tomato, and Aloe vera. Additionally, leveraging the LightGBM architecture, we demonstrate that MLIPBS successfully classifies combined stress conditions and varying intensity levels of acid and salt stress, achieving an average accuracy of 90.5%. We further show that the system identifies stress types and intensities within 8 hours of onset, providing an early-warning window at least 48 hours before symptom manifestation. Our study provides reliable wearable tools for stress-resistant crop screening and precision management in smart agriculture.
Accurate and non-destructive prediction of lettuce quality traits is essential for variety identification, germplasm utilization, and intelligent breeding. However, existing approaches relying on handcrafted features or purely data-driven models face limitations under small-sample conditions, including constrained prediction accuracy, weak interpretability, and an increased risk of overfitting. To address these challenges, we propose a knowledge-guided feature tokenizer transformer (KG-FT-Transformer) framework for hyperspectral quality trait prediction and fingerprint analysis. This framework integrates domain prior knowledge with data-driven learning, significantly improving prediction accuracy while enhancing biological interpretability. The KG-FT-Transformer employs a Transformer-based architecture integrating multi-head attention (MHA) with a gated feed-forward network (GFFN), enabling nonlinear spectral modeling and rich feature interactions. We evaluated its performance on three key quality traits: relative chlorophyll content (SPAD), soluble solids content (SSC), and moisture content (MC). The model achieved R 2 values of 0.9534, 0.9185, and 0.9226, with corresponding residual predictive deviation (RPD) values of 4.63, 3.50, and 3.60, outperforming all baseline models and demonstrating stable and consistent prediction performance. Moreover, pixel-wise predictions were used to construct quality trait fingerprints through pseudo-color mapping, intuitively visualizing the spatial distribution and varietal specificity of traits. SPAD and SSC exhibited visually consistent central aggregation patterns, while MC revealed distinct spatial variations among cultivars. These fingerprint-based representations provide spatially informed, qualitative references that may assist the interpretation of DUS-related (Distinctness, Uniformity, Stability) trait characteristics. Overall, this study demonstrates the potential of integrating hyperspectral prediction with quality fingerprinting for non-destructive quality assessment and breeding-oriented analysis, and provides a complementary perspective for germplasm identification and crop improvement.
• Multisensor platform integrates RGB, depth, IR, and RTK-GPS data streams • Automated plant segmentation and 3D reconstruction extract plant traits in field conditions • System validation shows high correlation with manual and lab measurements • Public RGB-D lettuce dataset released to support reproducible AI research Accurate monitoring of leafy vegetable crops is essential to evaluate plant health, growth, yield, and quality, yet conventional methods based on manual measurements are labor-intensive and error-prone. This study proposes a data-driven framework for automated in-field monitoring of a lettuce crop based on multidimensional data acquired by a ground platform under various field conditions. Specifically, an advanced perception system is developed, including imaging and localization sensors to capture high-resolution visual, structural, and georeferenced information on the crop. An image processing pipeline is then proposed using zero-shot learning for plant segmentation, followed by 3D phenotyping techniques based upon computational geometry to automatically estimate plant biophysical traits, thus minimizing human input. An experimental trial conducted in a test field in Bari, Italy, between April and May 2025 validated the approach against manual and laboratory estimations. The results demonstrate strong correspondence between automated and reference measurements with a Pearson correlation coefficient r > 0.9 for key traits, confirming the potential of the framework. The influence of different nitrogen levels on the growing cycle is also evaluated, showing that the proposed system may provide a useful tool for decision support in lettuce crop monitoring and management.
Understanding how light dynamically regulates ascorbic acid (AA) levels is essential for improving crop nutritional quality. However, the dynamic regulation of AA by light in vivo remains unclear, since conventional methods rely on destructive sampling and only provide static data. To address this, we develop an implantable fiber sensor functionalized with a dual-atomic nanozyme for minimally invasive, long-term tracking of AA in living plants. The implantable fiber sensor integrates a hierarchical nanobio interface composed of a Co-Fe dual-atomic nanozyme (CoFe-DAzyme) and an antifouling hydrogel, achieving a detection limit of 0.081 μM in plant bleeding sap and remaining functional for up to 7 days postimplantation. Employing this sensor in lettuce, we uncover rapid, light-dependent AA fluctuations, directly revealing how dynamic light environments fine-tune this key nutritional metabolite. Our work not only establishes a versatile sensing platform but also provides direct mechanistic insight into the light-regulated improvement of crop nutritional quality.
Aeroponic vertical tower farming is a cost-effective, sustainable method for optimizing the food crop-Lactuca Sativa (lettuce-a greeny leaf vegetable); yet accurate biomass prediction of the lettuce crop remains challenging due to the non-linear relationship between the climatic conditions and the variable lettuce growth parameters. To address this challenge, a robust machine learning model called UniTriRob regression model has been developed. This model primarily focuses on mitigating the effects of outliers and heteroskedastic errors across key growth-related parameters, including pH, total dissolved solids (TDS), temperature, electrical conductivity (EC), turbidity, humidity, light intensity and growth. The experimental validation highlights the model's capability with high R-squared value of 97.8386% and the minimized error rate of 0.46, that outperforms the conventional forecasting methods. Hence, the model presents a viable alternative for maximizing aeroponic lettuce production efficiency and increasing yield forecast accuracy, contributing to sustainable agricultural practices.
Computer vision and Internet of Things (IoT) technologies offer robust solutions for plant phenotyping, but traditional mainstream segmentation methods often fail in high-density plantings with overlapping foliage. This study introduces an integrated phenotyping system combining automated data capture and high-temporal RGB-D imaging using off-the-shelf hardware (Intel RealSense D435 and Raspberry Pi) to generate 3D point clouds of lettuce under controlled greenhouse conditions. While recent agricultural applications have shown limited success and required domain-specific adaptations, Segment Anything Model (SAM) and FastSAM were demonstrated to achieve exceptional zero-shot segmentation performance for individual lettuce plants in high-density arrangements without additional training. This capability effectively addresses the traditional challenges of species-specific parameter tuning and extensive training data requirements and fine-tuning. By mapping 2D segmentation masks to corresponding 3D point clouds, the system accurately extracted key phenotypic traits, namely plant height, length, and width, from which area and volume were subsequently estimated, showing strong correlations with manual measurements for Rex and Rouxai lettuce cultivars. This high-temporal, non-destructive monitoring provided unique insights into plant growth dynamics. The study highlights distinct growth patterns among these cultivars, underscoring the importance of tailored phenotyping approaches to optimise crop management strategies. By addressing the limitations of existing phenotyping methods, this work advances precision agriculture technologies, offering a cost-effective and efficient solution for monitoring dynamic crop growth with potential applications across various crops and growing conditions.
Computer vision and Internet of Things (IoT) technologies offer robust solutions for plant phenotyping, but traditional mainstream segmentation methods often fail in high-density plantings with overlapping foliage. This study introduces an integrated phenotyping system combining automated data capture and high-temporal RGB-D imaging using off-the-shelf hardware (Intel RealSense D435 and Raspberry Pi) to generate 3D point clouds of lettuce under controlled greenhouse conditions. While recent agricultural applications have shown limited success and required domain-specific adaptations, Segment Anything Model (SAM) and FastSAM were demonstrated to achieve exceptional zero-shot segmentation performance for individual lettuce plants in high-density arrangements without additional training. This capability effectively addresses the traditional challenges of species-specific parameter tuning and extensive training data requirements and fine-tuning. By mapping 2D segmentation masks to corresponding 3D point clouds, the system accurately extracted key phenotypic traits, namely plant height, length, and width, from which area and volume were subsequently estimated, showing strong correlations with manual measurements for Rex and Rouxai lettuce cultivars. This high-temporal, non-destructive monitoring provided unique insights into plant growth dynamics. The study highlights distinct growth patterns among these cultivars, underscoring the importance of tailored phenotyping approaches to optimise crop management strategies. By addressing the limitations of existing phenotyping methods, this work advances precision agriculture technologies, offering a cost-effective and efficient solution for monitoring dynamic crop growth with potential applications across various crops and growing conditions.
Lettuce ( Lactuca sativa ) is an important field crop, but our understanding of its phenotypic variation and underlying genetics under natural field conditions remains limited, posing challenges for identifying effective crop breeding targets. Longitudinal hyperspectral phenotyping allows for non-invasive monitoring of crop performance under diverse agricultural conditions. In this study, we used hyperspectral imaging to assess the phenotypic variation of almost 200 different field-grown lettuce varieties, following the same plants from just after seedling- to flowering-stage. With automated image processing, we extracted a wide range of spectral phenotypes related to metabolite content, growth efficiency, and environmental stress responses, creating a multi-dimensional time-resolved data set. Principal component analysis (PCA) revealed the major axes of spectral variation over time, and highlighted differences in spectral patterns among lettuce genotypes. Integrating on-site weather data, we modelled G×E interactions of reflectance, revealing regions of the lettuce vegetation spectrum that are primarily shaped by genotype and/or environment. We estimated phenotypic plasticity in response to time, temperature and rainfall using best linear unbiased predictions (BLUPs), capturing genotype-specific developmental trajectories and responses to the environment. We used genome-wide association studies (GWAS) to identify quantitative trait loci (QTLs) of PC-based, single and BLUP-based phenotypes, disentangling the genetic architecture of spectral lettuce phenotypes from major axes of variation down to single wavelength spectral plasticity. These findings provide new insights into the genome-wide genetic regulation and dynamics of spectral phenotypes in field grown lettuce.
Reproduction assets found保存済みの本文根拠を更新済みルールで再検証し、公開資産1件を確認しました。Code · publicScripts used in this study can be found at https://github.com/SnoekLab/Hyperspec_Mehrem_etal_2025.Open asset ↗SnoekLab/Hyperspec_Mehrem_etal_2025pdf-page:9 lines:1-31Code / dataset availability confirmedEurope PMC · checked 5 Sept 2026
Rapid and accurate identification of DUS (Distinctness, Uniformity, and Stability) test traits in lettuce leaves is essential for advancing multi-omics-driven intelligent breeding. It also plays a critical role in germplasm protection and enhancing agricultural competitiveness. However, the phenotypic traits of lettuce leaves are highly diverse and complex due to both genotypic variation and environmental influences, posing significant challenges for precise DUS trait quantification. To address these challenges, we propose a high-precision phenotypic trait extraction pipeline and introduce an interpretable phenotypic fingerprinting framework for lettuce subgroup identification. First, a lightweight semantic segmentation network guided by group attention is developed to extract leaf components. Then, shape, color, and texture traits are comprehensively quantified. Following UPOV (International Union for the Protection of New Varieties of Plants) guidelines, we establish quantitative methods for seven DUS test traits: leaf shape, leaf tip shape, leaf margin shape, leaf vein shape, color hue, brightness, and anthocyanin coloration. Finally, PCA (Principal component analysis) was used to select 13 key traits, capturing over 95.82% of the total variance, for constructing "phenotypic ID" of lettuce varieties. Experiments conducted on 709 lettuce leaf image datasets showed that the accuracy of subgroup identification based on phenotypic fingerprints reached 98.59%. This study offers a scalable approach for automated DUS test trait evaluation and intelligent crop variety identification, providing a novel paradigm with strong potential for application in precision breeding and germplasm resource management.
Reproduction assets foundThe article provides a public GitHub repository containing the authors' source code for the lettuce phenotypic fingerprint pipeline. The 709-image dataset and annotations are only available upon request, so they do not qualify as public assets.Code · publicThe data used to support the findings of this study are available upon request from the corresponding author, and the source code is accessible at https://github.com/qiuguangjie87/PP_Phenotypic_Fingerprint .Open asset ↗PP_Phenotypic_Fingerprintlines:263-278Plant phenotyping relevance matchEurope PMC · checked 15 Sept 2026
Published1 Mar 2026Computers and Electronics in Agriculture.
Accurate acquisition of phenotypic characteristics in protected crops is a crucial prerequisite for intelligent control and digital breeding in greenhouses. To accurately assess the phenotypic traits of protected lettuce, a specialized in situ phenotypic detection method has been developed. The Multimodal Features and Attention Mechanism for Phenotype Detection Model (MFAMNet) was developed for protected lettuce, employing a segmented multi-source image dataset for synchronous regression testing. The results revealed that the predicted values generated by MFAMNet exhibited a strong correlation with the measured values, achieving coefficients of determination of 0.96, 0.92, 0.95, 0.94, and 0.95 for plant height, crown width, leaf area, fresh weight, and dry weight, respectively. Ablation tests demonstrated that the deep learning detection framework based on multi-modal feature fusion significantly outperformed single-feature detection models, highlighting the advantages of integrating diverse data modalities. In addition, the multi-modal feature attention mechanism (MMF) facilitates both inter-modality and intra-modality interactions by capturing the global correlations between modalities and employing dynamic sparse spatial attention. The effectiveness of MMF has been validated through comparative experiments, demonstrating its suitability for the phenotypic detection of artificially cultivated lettuce. In summary, the method proposed in this study facilitates real-time monitoring of facility crops, enabling precise control of environmental parameters in protected agriculture and optimizing resource allocation. This approach contributes to the development of a comprehensive intelligent agriculture system and establishes a foundation for unmanned farms.
The integration of electronic system into agricultural production can significantly enhance its efficiency and scalability. However, most of the current research focuses on the data acquisition and automated control. The development of expert-level, interpretable decision-making systems remains a challenge, primarily due to the prohibitive requirement for extensive domain-specific labeled data. In this manuscript, a novel agentic framework integrated with Large Language Models is proposed and demonstrated, using seedling assessment as a case study. The framework achieves high predictive accuracy, strong interpretability, and fast-adaption ability, offering a distinct advantage over methods that demand large labeled datasets. An agentic orchestration framework integrated with the Analytic Hierarchy Process and the reasoning ability of the Large Language Models is constructed to automatically derive the raw assessment rating. Based on a score calibration system using few-shot learning with three different types of lettuce, Butterhead, Grand Rapids, and Ramosa Hort, the final rating score can be derived with good prediction accuracy based on a small dataset (less than 20 labelled data). Additionally, three supplementary plant species (Sprout, Ball Brassica, and Rapa Brassica) are used to demonstrate the framework’s rapid adaptation capability. A field experiment guided by the agentic framework is conducted to prove that this seedling assessment system can be applied to help increase yield by more than 20 %. Our framework presents an important attempt towards an intelligent agricultural system that is capable to achieve expert-level and data-efficient decision making, thereby helping to bridge the critical gap between artificial intelligence research and practical agricultural application.
The canopy characteristics of crops are essential aspects for assessing crop growth status and conducting phenotype analysis. As one of the key indicators to measure crop growth situation, accurate canopy coverage assessment can provide a strong foundation for crop growth and yield monitoring. Considering plant growth differences, this study investigated the statistical method for assessing canopy coverage using visual technology, focusing on lettuce as the research subject. Firstly, a multi-variety and multi-growth stage hydroponic lettuce image dataset was constructed, which lays a data foundation for the construction of a semantic segmentation model. Secondly, in order to ensure the precision of semantic segmentation, this study proposed a Channel-Axial-Spatial attention mechanism module from the perspective of feature enhancement. To satisfy the lightweight demands of practical model deployment, this study replaced the original backbone network of PSPNet with MobileNetv3, greatly reduced model complexity while minimizing model performance degradation. Finally, we developed a group lettuce canopy coverage acquisition system by employing Python in conjunction with PyQt5 and embedded the pre-trained models CAS-PSPNet and MobileNetv3-PSPNet into the system for effectiveness verification. By integrating the proposed attention mechanism module with PSPNet, the integrated model outperformed FCN, Unet, SegNet, Deeplabv3+, GCN, ExFusion, ENet, BiseNet, FusionNet, LinkNet, RefineNet, LWRefineNet, and PSPNet in semantic segmentation of lettuce plant groups, achieving a Mean Intersection over Union of 0.9832. The Mean Intersection over Union of PSPNet based on lightweight improvement is 0.9717, and the model size is 9.3M. The results show that the proposed semantic segmentation method can accurately capture the crop canopy coverage, offering a feasible solution for real-time crop growth monitoring.
Sustainable plant cultivation is critical for supporting long-duration space missions by ensuring reliable food production in extraterrestrial environments where resources are severely limited and growth systems operate in closed-loop conditions. With crew members managing multiple critical mission tasks and having minimal time for plant care, autonomous stress detection systems must provide reliable, interpretable diagnostics to enable rapid, informed decision-making for crop management. This study utilized a custom hyperspectral imaging (HSI) system designed for space applications to develop an AI-driven diagnostic framework. We propose a novel SAM-ViT-3PE architecture that uniquely combines sparse spectral band selection with 3D spatial-spectral patch embedding, preserving rich spatial-spectral information typically lost in conventional ROI-averaged approaches. A key temporal finding identified Day 3 After Treatment (DAT 3) as the critical threshold where drought stress signatures become distinctly detectable, with accuracy dramatically improving from 72.2% to 95.9%. By focusing analysis on data from DAT 3 onward, the SAM-ViT-3PE model achieved superior performance compared to traditional ML methods and standard deep learning approaches, with accuracy of 95.4%, precision of 96.6% and recall of 94.1%. Furthermore, Explainable AI using Integrated Gradients enabled interpretable diagnostics through physiologically meaningful spectral bands and spatial stress patterns. These results demonstrate that the AI-enhanced HSI framework provides both high-accuracy autonomous detection and scientifically grounded interpretability essential for trustworthy crop management in resource-constrained space environments.
Vertical farming offers a promising solution to global food security and urbanization challenges, yet its widespread adoption is hindered by high costs, particularly for lighting. Addressing this requires enhancing light use efficiency (LUE) through intelligent control strategies. While numerous studies have investigated the effects of light intensity on lettuce growth, relatively few have explored the potential benefits of stage-specific light regulation. In this study, we first developed an automated 3D phenotyping pipeline based on multi-view reconstruction to quantify canopy morphology and light interception. Utilizing this quantitative framework, we conducted a dynamic light experiment with lettuce in a commercial plant factory to evaluate four dynamic light-intensity strategies. The proposed 3D phenotyping pipeline demonstrated promising performance for canopy information extraction, with RMSEs for plant height, canopy diameter, and projected leaf area of 0.79 cm, 1.05 cm, and 44.3 cm², respectively. The “high-low-high” dynamic lighting strategy, applying higher light intensity during the early and late growth stages and lower intensity during the mid-growth stage, successfully optimized canopy morphology for better light capture. This treatment significantly increased shoot fresh and dry weights by 28 % and 65 %, respectively, compared to constant lighting. Furthermore, it enhanced LUE based on incident and intercepted light integrals by 67 % and 19 %, while reducing electricity consumption per unit of fresh weight by 24 %. Nutritional quality analysis showed the treatment increased soluble sugars and starch contents. By integrating advanced 3D phenotyping with dynamic light intensity control, this study demonstrates a prototype for intelligent decision-making to enhance yield and energy use efficiency in practical vertical farming.
LettuceTissuePhysiological trait estimationWater status / transpiration
With the rapid development of smart agriculture, the agricultural Internet of Things (Ag-IoT) has gradually established a monitoring system centered on distributed sensing, low-power communication, and intelligent control. However, current solutions underexplore the perceptible characteristics of communication signals, and therefore do not fully utilize their latent potential in environmental perception. This paper targets the demand for crop water monitoring and introduces an integrated sensing and communication (ISAC) approach. This method can achieve non-contact and continuous perception of crop water status by reusing the communication link without altering the existing hardware architecture and frequency band configuration. Taking leafy vegetables such as lettuce as the research object, a prototype system based on a 3 GHz communication link was built. A quantitative mapping model between the water content of plant tissues and the amplitude and phase disturbances they cause to electromagnetic waves was established. A joint optimization mechanism that considers both communication performance and sensing accuracy was proposed to achieve a coordinated configuration between communication quality (BER < 10⁻⁴, SNR ≈ 20 dB, EVM < 8 %) and sensing accuracy (MAE = 2.51 %, R² = 0.92). Experiments were conducted in controlled environments and production-like scenarios, demonstrating that the method can stably identify the water status of lettuce while ensuring communication quality of service (QoS). The proposed ISAC method is potentially compatible with existing Ag-IoT frequency bands and physical-layer infrastructures, assuming access to pilot/CSI and airtime control. Protocol-level integration with LoRa, Wi-Fi, and NB-IoT is defined as future work It provides a low-cost, high-integration, and easily scalable communication-driven solution for water monitoring in smart agriculture.
Plant motion provides valuable indicators of physiological responses to water stress. In this study, we present a structured image-based approach to define and test morpho-kinematic (MK) traits from lettuce plants subjected to varying irrigation regimes under controlled conditions. Four water availability treatments were imposed − Full Control (FC), Stress Control (SC), Mild Stress (SM), and Severe Stress (SS) − varying in timing, frequency, and intensity of irrigation protocols. Using dense optical flow on time-lapse RGB images, we extracted MK features that link leaf age to motion dynamics. These high-dimensional temporal features were compressed into descriptive and trend-based characteristics for classification. Multi-classification problem was divided into nine sub-tasks, for which feature selection and multiple machine-learning models were tested applying Leave-One-Sample-Out cross-validation. The best models were organised into four explainable hierarchical cascades. The presented system captures enough information to successfully distinguish among subtle differences in plants’ response to water availability dynamics (best architecture cascade obtained 0.93 out of fold balanced accuracy). The framework associating leaf age with MK features along with feature engineering allowed explainability – e.g., central rosette’s features were selected almost twice the expected frequency (19 out of 58) in tasks involving the stress-adapted control (SC), while features capturing linear trends in motion were generally selected over twice as often as simple descriptive statistics (44 vs. 19), proving essential for distinguishing most stress conditions. The MK approach proved effective for differentiating water stress levels, positioning it as a powerful tool for digital phenotyping and a solid foundation for developing advanced temporal-aware models.
Introduction: Disease detection in lettuce (Lactuca sativa L.) is crucial to enhance crop yields and prevent losses caused by bacterial, fungal, and weed-related infections. This study aimed to develop an Android-based lettuce disease detection application using a Convolutional Neural Network (CNN) algorithm to assist farmers in identifying plant diseases in real time. Method: The research used a dataset of 2,320 lettuce leaf images obtained from Kaggle, categorized as healthy, bacterial, fungal, and shepherd’s purse weed. The dataset was preprocessed through labeling, normalization, and augmentation to improve model robustness. The CNN architecture comprised four convolution layers followed by max-pooling, dense, and softmax output layers. The model was trained using TensorFlow and deployed through TensorFlow Lite for mobile implementation. Results: The CNN model achieved 93,67 % training accuracy and 93,99 % validation accuracy, demonstrating good generalization without overfitting. The evaluation using confusion matrix and classification reports showed high performance, particularly in identifying healthy and shepherd’s purse weed categories with F1-scores of 0.94 and 0.99, respectively. The Android application successfully detected diseases in real time and provided users with diagnostic results, historical data, and treatment suggestions. Conclusions: The developed CNN-based Android application proved effective for automatic lettuce disease detection with high accuracy and practical usability for farmers. Future studies could enhance performance through more advanced CNN architectures such as VGG16 or ResNet50 and the use of more detailed datasets for improved disease classification.
Biofeedback control of light‐emitting diode (LED) lighting based on real‐time photosynthetic performance offers a promising framework for plant‐responsive light management in controlled environment agriculture (CEA). While the short‐term feasibility of electron transport rate (ETR)‐based light regulation has been demonstrated, its long‐term performance remains untested. This study evaluated the ETR‐based biofeedback lighting control system over an entire crop cycle of lettuce under three target ETR levels (55, 90, and 125 μmol m⁻² s⁻¹) in a climate‐controlled growth chamber. The system continuously monitored the quantum yield of photosystem II (ΦPSII) and adjusted photosynthetic photon flux density (PPFD) every 15 min to maintain the target ETR, used as an indirect proxy for carbon assimilation. Target ETRs were maintained within ±2.5% with minimal variability among replicates, and the corresponding average PPFDs (means ± standard deviations) were 183.5 ± 5.4, 316.1 ± 14.3, and 457.3 ± 23.5 μmol m⁻² s⁻¹, respectively. Despite stable environmental conditions, the system dynamically responded to both diurnal and long‐term acclimation in terms of photosynthetic efficiency. PPFD was reduced during the early photoperiod, when ΦPSII was high, and increased in the late photoperiod to compensate for the decline in ΦPSII. Under the target ETR of 125 μmol m⁻² s⁻¹, ΦPSII increased over time, enabling a 14% reduction in PPFD while maintaining a stable ETR, highlighting the potential for reduced light input as plants acclimated. These results demonstrate the long‐term feasibility and stability of plant‐responsive, CF‐based biofeedback lighting control for precise and replicable regulation of photochemical energy input in CEA crop production.
Biofeedback control of light-emitting diode (LED) lighting based on real-time photosynthetic performance offers a promising framework for plant-responsive light management in controlled environment agriculture (CEA). While the short-term feasibility of electron transport rate (ETR)-based light regulation has been demonstrated, its long-term performance remains untested. This study evaluated the ETR-based biofeedback lighting control system over an entire crop cycle of lettuce under three target ETR levels (55, 90, and 125 μmol m -2 s -1 ) in a climate-controlled growth chamber. The system continuously monitored the quantum yield of photosystem II (Φ PSII ) and adjusted photosynthetic photon flux density (PPFD) every 15 min to maintain the target ETR, used as an indirect proxy for carbon assimilation. Target ETRs were maintained within ±2.5% with minimal variability among replicates, and the corresponding average PPFDs (means ± standard deviations) were 183.5 ± 5.4, 316.1 ± 14.3, and 457.3 ± 23.5 μmol m -2 s -1 , respectively. Despite stable environmental conditions, the system dynamically responded to both diurnal and long-term acclimation in terms of photosynthetic efficiency. PPFD was reduced during the early photoperiod, when Φ PSII was high, and increased in the late photoperiod to compensate for the decline in Φ PSII . Under the target ETR of 125 μmol m -2 s -1 , Φ PSII increased over time, enabling a 14% reduction in PPFD while maintaining a stable ETR, highlighting the potential for reduced light input as plants acclimated. These results demonstrate the long-term feasibility and stability of plant-responsive, CF-based biofeedback lighting control for precise and replicable regulation of photochemical energy input in CEA crop production.
Oswaldo Menéndez-Granizo · Alexis Chugá-Portilla · Tito Arevalo-Ramirez · Juan Pablo Vásconez · Fernando Auat-Cheein · Álvaro Prado-Romo
LettuceSugar beetGreenhouseLeafPhysiological trait estimationBiomass / plant weightWater status / transpiration
Large-scale wireless sensor networks with electric field energy harvesters (EFEHs) offer self-powered, eco-friendly, and scalable crop monitoring in hydroponic greenhouses. However, their practical adoption is limited by the low power density of current EFEHs, which restricts the reliable operation of external sensors. To address this challenge, this work presents a noninvasive EFEH assembled with hydroponic leafy vegetables that harvests electric field energy and estimates plant functional traits directly from the electrical response. The device operates through electrostatic induction produced by an external alternating electric field, which induces surface charge redistribution on the leaf. These charges are conducted through an external load, generating an AC voltage whose amplitude depends on the dielectric properties of the leaf. A low-voltage prototype was designed, built, and evaluated under controlled electric field conditions. Two representative species, Beta vulgaris (chard) and Lactuca sativa (lettuce), were electrically characterized by measuring the open-circuit voltage (VOC) and short-circuit current (ISC) of EFEHs. Three regression models were developed to determine the relationship between foliar moisture content (FMC) and fresh mass with electrical parameters. Empirical results disclose that the plant functional traits are critical predictors of the electrical output of EFEHs, achieving coefficients of determination of R2=0.697 and R2=0.794 for each species, respectively. These findings demonstrate that EFEHs can serve as self-powered, noninvasive indicators of plant physiological state in living leafy vegetable crops.
Introduction. Chlorophyll plays a crucial role in absorbing and transforming light energy into a chemical form that provides organic matter production in plants. Monitoring of chlorophyll content helps to assess plant-environment interactions and the degree of influence of stress factors that are essential for yield management. Traditional laboratory methods of analyzing are time-consuming, destroying samples and unsuitable for rapid field evaluations. A more reasonable solution is to use lowcost, portable devices. Aim of the Study. The study is aimed at developing and training an ANN architecture to predict the chlorophyll content in plant leaves based on their optical density within specific visible spectrum ranges. Materials and Methods. The artificial neural network dataset was compiled from experi- mental measurements using the DP-1M densitometer and the CCM-200 chlorophyll meter. Data were collected from lettuce, pepper, tomato and zucchini leaves of different ages, which were grown in different light environments. The artificial neural network training was carried out in the Google Colab environment with subsequent adaptation of the model for using in a microcontroller device – a photocolorimeter for leaves. Results. The dataset with 1,000 entries showed that the leaf optical density range isfrom 0.57 to 2.54 relative units (red), from 0.9 to 1.66 relative units (green), and from 1.09 to 3.53 relative units (blue). According to these data, the chlorophyll content variations are from 3.1 to 156.5 relative units. In the study, there were compared six artificial neural network architectures that differed by hidden-layer neurons. The structure “32:32” had the highest accuracy (MAE = 6.64 rel. units, MAPE = 16.34%, R² = 0.8886). A simplified structure “4:4” was selected to simplify the model and improve the microcontroller efficiency. This structure maintained the performance (MAE = 6.83 rel. units, MAPE = 16.86%, R² = 0.8808) with much smaller amount of resources used – 41 weight parameters and 164 bytes of memory. A comparative evaluation with classical machine learning algorithms demonstrated the superiority of the developed model across all metrics. Discussion and Conclusion. The trained artificial neural network was implemented on a microcontroller-based photocolorimeter for leaves that enabled the non-destroying optical density measurements. The developed model allows implementing non-destroying and operational monitoring of the condition of plants, which is especially important in precision farming systems. This approach has significant potential for ecological monitoring and precision agriculture. The study results demonstrate the viability of machine learning for improving plant status assessment and developing digital agrotechnology solutions.
Climate change is driving urgent demand for resilient crop varieties capable of withstanding extreme and changing conditions. Identifying resilient varieties requires systematic plant phenotyping research under controlled conditions, where dynamic environmental impacts can be studied. Current growth cabinets (GC) provide this capability but remain limited by high costs, static environments, and scalability. These limitations pose a challenge for climate change-based phenotyping research which requires large-scale trials under a variety of dynamic climate conditions. Presented is a microclimate-controlled smart growth cabinet (MCSGC) platform, addressing these limitations through four innovations. The first is dynamic microclimate simulation through programmable environmental ‘recipes’ reproducing real climactic variability. The second is interconnected scalable multi-cabinet for parallel experiments. The third is modular hardware able to reconfigure for different plant species, remaining cost-effective at <$10,000 AUD. The fourth is automated data collection and synchronisation of environmental and phenotypic measurements for Artificial Intelligence (AI) applications. Experimental validation confirmed precise climate control, broad crop compatibility, and high-throughput data generation. Environmental control stayed within ±2 °C for 97.42% while dynamically simulating Hobart, Australia, weather. The MCSGC provides an environment suitable for diverse crops (temperature 14.6–31.04 °C, and Photosynthetically Active Radiation (PAR) 0–1241 µmol·m−2·s−1). Multi-species cultivation validated the adaptability of the MCSGC across Cannabis sativa (544.1 mm growth over 34 days), Beta vulgaris (123.6 mm growth over 36 days), and Lactuca sativa (19-day cultivation). Without manual intervention the system generated 456 images and 164,160 sensor readings, creating datasets optimised for AI and digital twin applications. The MCSGC addresses critical limitations of existing systems, supporting advancements in plant phenotyping, crop improvement, and climate resilience research.
LettuceGrowth chamberRGB / grayscaleLeafClassificationWater status / transpiration
Plant motion provides valuable indicators of physiological responses to water stress. In this study, we present a structured image-based approach to define and test morpho-kinematic (MK) traits from lettuce plants subjected to varying irrigation regimes under controlled conditions. Four water availability treatments were imposed − Full Control (FC), Stress Control (SC), Mild Stress (SM), and Severe Stress (SS) − varying in timing, frequency, and intensity of irrigation protocols. Using dense optical flow on time-lapse RGB images, we extracted MK features that link leaf age to motion dynamics. These high-dimensional temporal features were compressed into descriptive and trend-based characteristics for classification. Multi-classification problem was divided into nine sub-tasks, for which feature selection and multiple machine-learning models were tested applying Leave-One-Sample-Out cross-validation. The best models were organised into four explainable hierarchical cascades. The presented system captures enough information to successfully distinguish among subtle differences in plants’ response to water availability dynamics (best architecture cascade obtained 0.93 out of fold balanced accuracy). The framework associating leaf age with MK features along with feature engineering allowed explainability – e.g., central rosette’s features were selected almost twice the expected frequency (19 out of 58) in tasks involving the stress-adapted control (SC), while features capturing linear trends in motion were generally selected over twice as often as simple descriptive statistics (44 vs. 19), proving essential for distinguishing most stress conditions. The MK approach proved effective for differentiating water stress levels, positioning it as a powerful tool for digital phenotyping and a solid foundation for developing advanced temporal-aware models.
In situ detection of plant ion signals faces technical limitations in terms of real-time capability, minimal invasiveness, and data analysis. Therefore, the development of sensors for in vivo plant detection and construction of time-series prediction models to analyze the dynamic patterns of ion concentrations in plants are imperative. This study presents a microneedle electrode system for potassium ion (K⁺) sensing, which is applied to real-time in situ detection in lettuce. The microneedle ion-selective electrodes (ISEs) fabricated herein exhibited a rapid potentiometric response (within < 15 s), with concentration responses adhering to the Nernst equation. During in vivo plant detection, the system captured instantaneous ion-signal changes upon exogenous application without influencing subsequent plant growth. This study demonstrates the pioneering application of time-series prediction (nonlinear autoregressive neural network model) to analyze in vivo K⁺ signals in lettuce, accurately forecasting ion concentration dynamics over time and identifying the transition pattern from signal fluctuation to stabilization. The integration of microneedle ISE-based in situ plant monitoring with time-series prediction represents a crucial and reliable approach to agricultural sensor innovation, providing a novel paradigm for precision agriculture and plant stress response research.
Lettuce ( Lactuca sativa ), a widely cultivated leafy vegetable, is highly susceptible to bacterial and fungal infections that severely reduce yield and quality. Rapid and accurate disease identification is therefore essential for precision agriculture and sustainable crop management. This study proposes Efficient-FBM-FRMNet, a modular deep learning framework for automated lettuce disease detection. The model integrates EfficientNetB4 with dilated convolutions, a Feature Bottleneck Module (FBM) for redundancy reduction, a Reasoning Engine for higher-order semantic inference, and a Feature Refinement Module (FRM) for enhanced generalization. The framework was trained and validated on a publicly available dataset of 2,813 lettuce leaf images (bacterial, fungal, and healthy classes) using stratified 5-fold cross-validation. The proposed Efficient-FBM-FRMNet achieved an overall accuracy of 97.5%, outperforming baseline CNNs such as EfficientNetB4, ResNet50, and DenseNet121. It demonstrated superior precision (96.0%), recall (96.6%), and F1-score (97.0%), confirming its robustness and consistency across multiple folds. Statistical significance analysis (p
Reproduction assets foundThe paper's phenotyping measurements are based entirely on a public Kaggle lettuce plant disease image dataset (2,813 images, bacterial/fungal/healthy), cited with an explicit public URL matching an allowed URL. No author code or model checkpoints are stated as available.Dataset · publiceelwal P.
Dhiman P.
Gulzar Y.
Kaur A.
Wadhwa S.
Onn C. W.
( 2024 ).
A systematic review of deep learning applications for rice disease diagnosis: current trends and future directions
. Front. Comput. Sci.
6 . doi:
10.3389/fcomp.2024.1452961
Shaha, S.
.(n.d.) Lettuce plant Disease Dataset [Data set]. Kaggle. Available online at: https://www.kaggle.com/datasets/santoshshaha/lettuce-plant-disease-dataset (Accessed March 12, 2025 ).
Shoaib M. A.
Lai K. W.
Chuah J. H.
Hum Y. C.
Ali R.
Dhanalakshmi S.
. ( 2022 ).
Comparative studies of deep learning segmentation models for left ventricle segmentation
. Front. Public Health
10 , 981019 . doi:
10.3389/fpubh.2022.981019
, PMID:
36091529
PMC9453312
SuOpen asset ↗Kaggle · lettuce-plant-disease-datasetlines:990-1174Plant phenotyping relevance matchEurope PMC · checked 6 Sept 2026
Abstract Background and Aims: Portable X-ray fluorescence spectrometry (pXRF) has emerged as a robust analytical approach for elemental determination in plant tissues, enabling rapid, non-destructive, and reagent-free measurements. This study developed and validated an empirical calibration of pXRF for quantifying silicon (Si) in plants, using autoclave-induced digestion (AID) as the reference method. Methods A total of 374 samples from seven plant species (rice, maize, soybean, cowpea, sorghum, lettuce, and beet) were analyzed. Silicon concentrations obtained via AID ranged from 1.07 to 19.23 g kg − ¹ (mean = 4.48 g kg − ¹; coefficient of variation = 67%), reflecting substantial interspecific variability. Each sample was also analyzed by pXRF under optimized instrumental conditions, and a calibration model was constructed using 75% of the dataset to predict Si concentrations relative to AID values. Results The pXRF calibration exhibited a strong linear relationship with AID results (R² = 0.94; R = 0.97; p
This study investigates lettuce growth under extreme environmental conditions by simulating the weather in six climate zones in a plant growth chamber, including Lleida, Adelaide, Paris, San Luis, Singapore, and Fairbanks. The experiment involved weekly exposure to a new city’s climate, simulating “non-terrestrial weather stress,” which is also motivated from the vantage point of space plant growth and its process-control limitations. These simulated conditions shed light on ‘Climate 2050’, when Earth will probably have harsher and more fluctuating conditions. For the period investigated, the real temperature changes could be reproduced well and in real-time in the growth chamber, the actual rain fall was mimicked, and the lighting period was adjusted to the real sunshine exposure in the respective city. The virtual move of the lettuce plant from between six climates with their own profile in temperature, lighting time, and water is assumed to create stress beyond the variability of a weather change within a single climate. Machine learning models, including linear regression, random forest regression, and boosted decision tree regression, were employed to predict weekly lettuce biomass and yield. This study successfully demonstrated the application of machine learning algorithms for predicting lettuce growth under the given range of six climate conditions. Among the tested models, random forest regression consistently delivered the most accurate and reliable biomass predictions, achieving an R² of nearly 99 % and MAPE of 6 % in all scenarios. By introducing tuned correction factors for conditions like drought stress, fertilisation, and mixed soil composition, the accuracy and flexibility of models are enhanced. This research highlights the value of integrating real-time data with machine learning through a digital twin framework, offering a promising direction for climate-resilient agriculture and space-based plant growth systems.
Vegetation indices (VIs) are a widely adopted and straightforward tool for non-contact estimation of chlorophyll and carotenoid content in plant leaves. However, VI-based method accuracy depends critically on instrument configuration and calibration procedures. This study aimed to evaluate the sensitivity of VI-based pigment assessment to variations in spectral channel parameters (central wavelength and bandwidth) as well as to changes in calibration details defined by the specific VI formula. Pigment content was measured in leaves of Lactuca sativa L. and Cucumis sativus L. at contrasting developmental stages using VI-based reflection spectroscopy across the 450-950 nm spectral range with various protocols and spectrophotometry as the reference method. VI values were calculated with varying central wavelength and widths of spectral bands, and across different VI formulas. Comparative analysis of the obtained measurements revealed that even minor shifts in central wavelengths of less than 20 nm or the use of an alternative index formula could lead to relative errors of 42-77% in the estimation of chlorophylls and carotenoids content, while changes in bandwidth had a much smaller impact, resulting in only 2-5% relative errors. Even with identical parameters of spectral channels, the choice of an appropriate VI and its regression model could introduce significant errors, ranging from 36% to 86%. These findings highlight the critical role of instrument specifications and calibration models in the VIs-based method accuracy and stability, as measurement errors can lead to suboptimal agronomic decisions. Moreover, our study underscores that comparing results from different sensors or platforms can be unreliable unless the channel parameters and calibration details are clearly specified. Therefore, standardization and transparency in VIs assignment is vital to ensure reproducibility and cross-compatibility in non-destructive pigment monitoring by using various devices.
Improving light-use efficiency (LUE) is essential for boosting crop productivity, particularly in controlled-environment agriculture. Despite recent advances, most studies still rely on destructive measurements or one-dimensional data, which limits insight into the structural–physiological coordination underlying LUE. We established a multimodal phenotyping platform to dissect the phenotypic regulatory network of LUE in lettuce ( Lactuca sativa L.). Integrating hyperspectral imaging with multiview three-dimensional (3D) reconstruction, we developed a noninvasive, high-throughput system that simultaneously estimates 3D plant architecture, photosynthetic physiology—net photosynthetic rate (A) and relative chlorophyll content (SPAD)—and aboveground biomass (AGB) across 35 cultivars. A modeling pipeline combining StandardScaler (SS) normalization, genetic algorithm (GA) feature selection, and artificial neural networks (ANN) achieved robust prediction of A (R²=0.72), SPAD (R²=0.87), and AGB (R²=0.85). Spectral contribution analysis revealed distinct sensitivities: SPAD across 400–700 nm, A near 430 and 680 nm, and AGB across 500–580 nm. The 426–430 nm blue band emerged as a key region: high-efficiency cultivars showed distinctive reflectance (42.93–59.03 %), consistent with superior photosynthetic performance. Structurally, high-efficiency types exhibited “large-and-loose” canopies, with greater plant height (+64.37 %), projected area (+59.42 %), and convex-hull volume (+166.3 %), alongside reduced compactness (−23.48 %). Network analysis indicated progressively tighter coupling between spectral and structural traits from low- to high-efficiency groups, consistent with adaptive coordination for light capture and use. These results identify actionable phenotypic markers for selecting high-LUE cultivars and provide a transferable platform for phenomics-driven breeding and management in controlled-environment crops. • A multimodal framework enables non-destructive, high-throughput phenotyping in lettuce. • 66 key spectral and structural features linked to light-use efficiency were identified. • A photosynthetic trait network reveals coordination of pigments and canopy architecture. • Breeding targets for blue-light response and canopy structure optimization are proposed.
Chen Z, Carter LJ, Banwart SA, Roychoudhry S, Pramanik DD, Kay P.
Brassica vegetablesLettuceRadishMicroscopyMultimodalRootVisualization / data management
Microfibers (MFs), primarily originating from sewage sludge and laundry effluents, are the most prevalent form of microplastics (MPs) in agricultural soils. While their ecological effects have been explored, the visualization, crop-level accumulation, and potential transport mechanisms of MFs within soil-plant systems remain poorly understood. This study combines 1,3,6,8-pyrene tetrasulfonic acid (PTSA) fluorescent staining with a sequential multimodal microscopy workflow to effectively track the distribution, adsorption, accumulation, and uptake of MFs under realistic soil cultivation conditions. Three edible vegetables-lettuce, Chinese cabbage, and cherry radish-were used to evaluate species-specific response patterns. The results revealed clear differences in MF interactions across species: lettuce exhibited strong MF adsorption on root surfaces and subsequent penetration via crack-entry and apoplastic pathways without entering cells. In contrast, Chinese cabbage and cherry radish showed limited MF adsorption and no uptake. These patterns were associated with root permeability and antioxidative capacities, indicating that plant functional traits play a critical role in determining the transport capacity of MPs. Beyond introducing a novel method for MF visualization in complex terrestrial matrices, this study provides new insights into the risks posed by MFs to soil-plant systems. The findings also highlight potential threats to food safety and underscore the need to establish plant-specific thresholds and pollution mitigation strategies to support sustainable agriculture and protect public health.
Plant pigment content is a crucial indicator for assessing photosynthetic efficiency, nutritional status, and physiological health. Its spatial distribution is significantly influenced by variety, location, and environmental factors. However, existing methods for measuring pigment content are often destructive, inefficient, and costly, making them unsuitable for the demands of modern precision agriculture. This study proposes a cross-scale, non-destructive detection method for lettuce pigments by integrating hyperspectral imaging (HSI) technology with deep learning algorithms, addressing the limitations of existing techniques in high-throughput and spatial resolution analysis. In this study, we built a multidimensional dataset based on eight different types of lettuce and developed a deep learning model named LPCNet to predict the contents of chlorophyll a (Chl a), chlorophyll b (Chl b), carotenoids (Car), and total pigment content (TPC) in lettuce. The LPCNet model integrates convolutional neural networks (CNN), bidirectional long short-term memory networks (BiLSTM), and multi-head self-attention (MHSA) mechanisms, enabling automatic extraction of pigment-related key features and simplifying the complex preprocessing and feature selection procedures required in traditional machine learning. Compared to multivariate analysis methods in machine learning, LPCNet demonstrated superior predictive accuracy, with coefficients of determination ( RP2 ) of 0.9449, 0.8613, 0.9121, and 0.8476 for Chl a, Chl b, Car, and TPC, respectively. Additionally, by combining the hyperspectral reflectance of lettuce canopies with the leaf-level inversion model, we visualized the spatial distribution of pigment content on the canopy of lettuce, achieving cross-scale analysis from leaf to canopy. This study provides an innovative approach for the rapid and accurate assessment of lettuce pigment content and offers an effective visualization tool for revealing the physiological processes and growth development of lettuce.
Reproduction assets foundThe article's data availability statement points to a public GitHub repository containing the authors' spectral analysis code for the LPCNet pigment-inversion workflow. No public phenotype dataset or hyperspectral image deposit is stated; supplementary data is only a small docx.Code · publicFurther details of the code are available at: https://github.com/zhaoyyy620/spectral_analysis.Open asset ↗zhaoyyy620/spectral_analysishtml-lines:448-472Plant phenotyping relevance matchOpenAlex · checked 13 Sept 2026
In situ detection of growth information in greenhouse crops is crucial for germplasm resource optimization and intelligent greenhouse management. To address the limitations of poor flexibility and low automation in traditional phenotyping platforms, this study developed a controlled environment inspection robot. By means of a SCARA robotic arm equipped with an information acquisition device consisting of an RGB camera, a depth camera, and an infrared thermal imager, high-throughput and in situ acquisition of lettuce phenotypic information can be achieved. Through semantic segmentation and point cloud reconstruction, 12 phenotypic parameters, such as lettuce plant height and crown width, were extracted from the acquired images as inputs for three machine learning models to predict fresh weight. By analyzing the training results, a Backpropagation Neural Network (BPNN) with an added feature dimension-increasing module (DE-BP) was proposed, achieving improved prediction accuracy. The R2 values for plant height, crown width, and fresh weight predictions were 0.85, 0.93, and 0.84, respectively, with RMSE values of 7 mm, 6 mm, and 8 g, respectively. This study achieved in situ, high-throughput acquisition of lettuce phenotypic information under controlled environmental conditions, providing a lightweight solution for crop phenotypic information analysis algorithms tailored for inspection tasks.
The identification of germplasm resources and analysis of phenotypic traits in lettuce hold significant importance for screening superior varieties and advancing genetic research. To uncover phenotypic differences among lettuce types and their dynamic changes during growth, this study utilized high-throughput phenotyping platform (HTPP) to collect and analyze time-series image data of eight lettuce types. Lettuce phenotypic traits were classified into five major categories: morphology, structure, color, texture, and size, which were further subdivided into 24 subcategories. Using multivariate statistical analysis, the study explored the relationships between phenotypic traits and lettuce types. Combining time-series analysis, the study examined the associations among phenotypic traits, lettuce types, and temporal sequences, revealing the dynamic changes of different lettuce types during their growth processes. The results showed that there were significant differences in phenotypic traits among different lettuce types throughout the growth cycle. These differences reflect the genetic characteristics and phenotypic variation patterns among lettuce types, revealing that genotype guides phenotype formation, while different phenotypes directly influence lettuce growth dynamics. Furthermore, by integrating multidimensional phenotypic traits, we constructed a phenotypic fingerprint for lettuce, providing each lettuce plant with a unique identifier that enables rapid detection of phenotypic differences among lettuce individuals and assists in the selection of elite cultivars. This study provides technical support for precise and rapid identification of lettuce germplasm resources, and can be used as the data basis for genetic research.
In vivo plant imaging is crucial for understanding plant biology and the influence of external factors on plant health. While exogenous contrast agents are widely used in animal bioimaging to enhance contrast, track flows, and provide molecular specificity, their application in plant tissue remains challenging and underexplored. Herein, we highlight the shortcomings of contrast agents in optical coherence tomography of plant tissue while demonstrating the successful use of Au bipyramids (AuBPs) as effective NIR-II contrast agents in photoacoustic imaging, enabling spatiotemporal flow tracking in live Buttercrunch lettuce over 5 days. Furthermore, rigorous plant health studies showed no adverse effects of the AuBPs on the physiological properties of Buttercrunch lettuce 7 days postinfiltration. The use of AuBPs as contrast agents in plant imaging, combined with the versatility of their surface chemistry, opens the possibilities for studies with molecular specificity.
Circadian clocks pace biological events and influence developmental traits, but their role in leafy crop domestication has remained unexplored. We investigated whether selection for delayed bolting during lettuce domestication targeted clock components. We phenotyped circadian rhythms and developmental timing across 234 cultivated and wild lettuce accessions. Using high-throughput leaf movement tracking, genome-wide association studies (GWAS) and haplotype analyses, we identified genetic variants controlling both the clock period and bolting time. Cultivated lettuce exhibits a significantly longer circadian period than its wild relatives, associated with a truncating mutation in PHYTOCHROME C (PHYC). This allele is not only associated with a decelerated clock but also with delayed bolting and flowering time in multiple field experiments. The truncating PHYC allele (H02) is enriched in modern cultivars and phylogenetically close to the wild ancestor (Lactuca serriola) alleles, indicating an early selection during lettuce domestication. Our study directly links for the first time circadian clock deceleration to domestication and breeding in a leafy crop. PHYC emerges as a pleiotropic regulator of the clock and developmental timing shaped by selecting delayed bolting during lettuce domestication and breeding. We demonstrate that circadian phenotyping is a powerful, scalable tool to predict developmental timing and uncover targets for crop improvement.
To address the problems of traditional methods that rely on destructive sampling, the poor adaptability of fixed equipment, and the susceptibility of single-view angle measurements to occlusions, a non-destructive and portable device for three-dimensional phenotyping and biomass detection in lettuce was developed. Based on the Structure-from-Motion Multi-View Stereo (SFM-MVS) algorithms, a high-precision three-dimensional point cloud model was reconstructed from multi-view RGB image sequences, and 12 phenotypic parameters, such as plant height, crown width, were accurately extracted. Through regression analyses of plant height, crown width, and crown height, and the R2 values were 0.98, 0.99, and 0.99, respectively, the RMSE values were 2.26 mm, 1.74 mm, and 1.69 mm, respectively. On this basis, four biomass prediction models were developed using Adaptive Boosting (AdaBoost), Support Vector Regression (SVR), Gradient Boosting Decision Tree (GBDT), and Random Forest Regression (RFR). The results indicated that the RFR model based on the projected convex hull area, point cloud convex hull surface area, and projected convex hull perimeter performed the best, with an R2 of 0.90, an RMSE of 2.63 g, and an RMSEn of 9.53%, indicating that the RFR was able to accurately simulate lettuce biomass. This research achieves three-dimensional reconstruction and accurate biomass prediction of facility lettuce, and provides a portable and lightweight solution for facility crop growth detection.
Rens Dijkhuizen · Abraham L. van Eijnatten · Sarah L. Mehrem · Esther S. van den Bergh · Jelmer van Lieshout · Kiki Spaninks · Steven Kaandorp · Remko Offringa · Marcel Proveniers · Guido Van den Ackerveken · Basten L. Snoek · Sub Bioinformatics · Sub Translational Plant Biology · Theoretical Biology and Bioinformatics · Translational Plant Biology
In recent years, accurate and low-cost variant calling has enabled the genotyping of large diversity panels for genome-wide association studies. As a result, phenotyping rather than genotyping is now the rate-limiting step, especially in field experiments. This has created a strong need for high-throughput, accurate, and low-cost in-field phenotyping. Here, we present a genome-wide association study (GWAS) study on 194 field-grown accessions of lettuce (Lactuca sativa). These accessions were non-destructively phenotyped at two time points 15 days apart using a drone equipped with an RGB and multispectral (MSP) camera. Our high-throughput phenotyping approach integrates an RGB- and MSP camera to measure the color and height of lettuce in this large-scale field experiment. We used the mean and other summary statistics, such as median, quantiles, skewness, kurtosis, minimum, and maximum to quantify different aspects of color and height variation in lettuce from the drone images. Using these summary statistics as traits for GWAS, we confirm several previously described genetic associations, now under field conditions, and identify additional novel associations for color and height traits in lettuce.
Rens Dijkhuizen · Abraham L. van Eijnatten · Sarah L. Mehrem · Esther S. van den Bergh · Jelmer van Lieshout · Kiki Spaninks · Steven Kaandorp · Remko Offringa · Marcel Proveniers · Guido Van den Ackerveken · Basten L. Snoek
In recent years, accurate and low-cost variant calling has enabled the genotyping of large diversity panels for genome-wide association studies. As a result, phenotyping rather than genotyping is now the rate-limiting step, especially in field experiments. This has created a strong need for high-throughput, accurate, and low-cost in-field phenotyping. Here, we present a genome-wide association study (GWAS) study on 194 field-grown accessions of lettuce (Lactuca sativa). These accessions were non-destructively phenotyped at two time points 15 days apart using a drone equipped with an RGB and multispectral (MSP) camera. Our high-throughput phenotyping approach integrates an RGB- and MSP camera to measure the color and height of lettuce in this large-scale field experiment. We used the mean and other summary statistics, such as median, quantiles, skewness, kurtosis, minimum, and maximum to quantify different aspects of color and height variation in lettuce from the drone images. Using these summary statistics as traits for GWAS, we confirm several previously described genetic associations, now under field conditions, and identify additional novel associations for color and height traits in lettuce.
Reproduction assets foundThe paper's authors publicly deposited their image processing, GWAS, and figure scripts on GitHub (SnoekLab/Dijkhuizen_etal_2025_Drone) and all raw/intermediate phenotyping data (including weather data) at a UU Yoda DOI (10.24416/UU01-S5FCM9). Both are paper-specific, public, and actionable.Code · publicThe scripts for making the SNP map from the filtered VCF file and for the image processing, GWAS, and figures in this manuscript are available on https://github.com/SnoekLab/Dijkhuizen_etal_2025_Drone .Open asset ↗SnoekLab/Dijkhuizen_etal_2025_Dronelines:362-531Dataset · publicData available at https://doi.org/10.24416/UU01‐S5FCM9 . This includes all raw data, all intermittent steps, the data required to generate all figures, and data on the weather during the experiment.Open asset ↗10.24416/UU01‐S5FCM9lines:362-531Plant phenotyping relevance matchEurope PMC · checked 6 Sept 2026
Studies on plant ecotoxicology focus essentially on growth and biochemical processes, often overlooking anatomical and morphological alterations that may occur post-germination. These changes, however, offer valuable insights into early environmental stress, enabling proactive intervention and mitigation strategies. Thus, this study aimed to identify and characterize morphological and anatomical alterations during early seedlings' development, using a new Visual PhytoToxicity assessment (ViPTox) approach. This visual scoring system categorizes the alterations recorded into severity levels, offering a simple, reproducible method for assessing phytotoxicity based on observable changes in plant structure. A standard germination assay with Lactuca sativa was conducted using potassium dichromate (PD, reference compound) at 0.00, 100.5, 120.6, 144.7, 173.6, 208.3, and 250.0 mg/L. Standard endpoints, including germination rate, seedling size, and fresh and dry weight were evaluated alongside a novel ecotoxicological approach. Based on the seedling effects observed, a dichotomous key with a scoring system was defined using a classification range from 0 (normal seedling) to 10 (no germination - maximum damage), providing insights into the severity of observed alterations (e.g., absence of roots and/or leaves (score 9), chlorosis and necrosis (score 8), atrophy (score 7, 6, and 5), deformations (score 4 and 3), reduction of size (score 2 and 1)), in order to calculate the phyto-morphological damage (PMD). Considering the standard endpoints, no significant alterations were observed in L. sativa germination. However, a significant decrease in seedling size (> 20 %) and fresh weight (> 50 %) was observed, after exposure to the highest PD concentrations (173.6, 208.3, and 250.0 mg/L). Regarding the ViPTox approach, PMD was observed in all concentrations ≥ 120.6 mg/L of PD. Significant effects were observed even at lower PD concentrations (120.6 and 144.7 mg/L) where phyto-morphological damages (e.g., atrophy and deformations) were quantified, while standard endpoints were unaffected. ViPTox presents a reproducible, non-invasive, and cost-effective approach to evaluate seedling responses to environmental stress, complementing traditional assessment techniques while providing crucial insights that support proactive intervention and effective mitigation strategies.
As the global population is expected to reach 10.3 billion by the mid-2080s, optimizing agricultural production and resource management is crucial. Climate change and environmental degradation further complicate these challenges, impacting crop productivity and food security. Traditional farming methods struggle with efficiently managing nutrients and water while ensuring high-quality products, leading to resource wastage and food safety concerns. This study aims to develop a hybrid model combining machine learning and physics-based techniques to predict fresh weight, leaf area, nitrate levels, and water consumption in lettuce grown in aeroponic systems, thereby enhancing resource management and product quality. We integrated a physics-based model with machine learning algorithms to create a dynamic hybrid framework. The model was validated with real-time data from aeroponic systems, showing good predictive performance, particularly for fresh weight and total leaf area. In contrast, predictions of nitrate content and water consumption were less accurate, due in part to smaller training datasets and limitations of the physics-based component under soilless conditions. Despite these challenges, the hybrid model offers a promising solution for optimizing controlled environment agriculture, addressing critical challenges in modern agriculture by improving efficiency and sustainability.
Accurate monitoring of nutrient supply is crucial for effective plant nutrition management. This study introduces a bioimpedance spectroscopy (BIS) approach to evaluate how lettuce genotypes respond to nitrogen (N) supply, illuminating the link between BIS and N metabolism. The research consisted of two parts: a preliminary experiment focusing on extracellular fluid resistance, followed by a second experiment involving five different N treatments applied to two lettuce genotypes. BIS spectra were recorded from 1 Hz to 100 kHz, with a total of 224 measurements. Plant physiological analyses assessed total and extracellular fluid nitrate (NO₃⁻) concentrations, pigment concentration, total N content, and cell membrane stability index. In the preliminary experiment, a significant negative correlation was observed between extracellular fluid NO₃⁻ concentration and extracellular fluid resistance (r = – 0.67). Furthermore, the frequency range around 63 Hz (R² = 0.97) appeared most sensitive to NO₃⁻ dynamics in the ‘Kobak’ lettuce genotype. In the two lettuce genotypes exposed to the five different N treatments, extracellular fluid resistance negatively correlated with total N and NO₃⁻ content (r = – 0.76 and – 0.73), while vacuole fluid resistance showed a moderate correlation (r = – 0.51). Cell membrane capacitance was also significantly correlated with membrane stability index (r = 0.68), indicating that low-N conditions reduce membrane charge storage. However, these correlations appeared highly genotype dependent. Heatmap-based hierarchical clustering of Z-score standardized BIS parameters further confirmed these genotype-specific patterns. Collectively, these findings suggest that BIS captures key aspects of N uptake and storage, particularly in the extracellular fluid compartment. Overall, BIS offers a robust, non-destructive method for monitoring N status and may enhance precision in nutrient management for lettuce and other crops, although the genotype-specific differences must be carefully considered.
Lettuce (Lactuca sativa) is the most consumed leafy vegetable in the world, with great economic and social importance in Brazil. In breeding programs, selecting genotypes with high agronomic potential is essential to meet market demands and cultivation conditions. In this context, plant phenotyping by means of multispectral imaging emerges as a modern, efficient and non-destructive tool, which enhances the analysis of phenotypic characteristics quickly and accurately. Therefore, the aim of the present study was to group different lettuce situations according to their group using image-based phenotyping, in addition to morphological descriptors and agronomic evaluations. The experiment was carried out in an experimental area of the Federal University of Uberlândia, Campus of Monte Carmelo, MG, Brazil, in randomized blocks with three replicates and 17 treatments (lettuce populations of the F2 generation, resulting from the cross between different lettuce cultivars and/or lines). Morphological descriptors and agronomic characteristics were obtained in the field. The vegetation indices GLI, NDVI, GNDVI, NGRDI and NDRE were calculated from images acquired at 49 days after transplanting. Means were compared using the Scott–Knott test (p ≤ 0.05), and the results were presented in box plots. Genetic dissimilarity was confirmed by multivariate analysis, which resulted in a cophenetic correlation coefficient of 96.11%. In addition, validation between field-collected data and image-obtained data was performed using heat maps and Pearson’s correlation. Populations UFU 003, UFU 006, UFU 009, UFU 011, UFU 012, UFU 013, UFU 014, UFU 016 and UFU 017 stood out, with high agronomic potential. Image-based phenotyping was correlated with agronomic traits and, therefore, can be considered an alternative to grouping different lettuce populations.
Climate change poses a major threat to crop production, resulting in the emergence of new pests and diseases. Phytophthora cryptogea has recently emerged as a major concern in hydroponic lettuce cultivation, causing substantial yield and economic losses. This oomycete pathogen thrives in elevated water temperatures induced by warmer weather conditions (e.g., heatwaves), facilitating rapid pathogen propagation. Although the disease is already present for several decades in chicory cultivation, where it originates from the field, its origin in lettuce cultivation remains unclear. To get a better understanding of its origin, we conducted a multilocus sequence analysis using five reference genes (ITS, β-tub , COI , EF1α , and HSP90 ) and 33 P. cryptogea isolates from various hosts, including chicory and lettuce. Results revealed a clear separation between lettuce and chicory isolates. Furthermore, we developed and implemented a robust disease bioassay and qPCR assay to investigate the interaction between P. cryptogea strains and lettuce. Our findings revealed that while lettuce isolates exhibited the highest virulence, some chicory isolates also caused disease in lettuce, suggesting a potential evolutionary link between P. cryptogea in lettuce and chicory. Our experiments also revealed that even a low concentration of zoospores (100 zoospores/liter) can elicit severe symptoms, underscoring the pathogen's high virulence. Therefore, effective disease management strategies are needed for controlling (the spread of) the disease. Together, this research provides several tools that can be used to enhance our understanding of the interaction between P. cryptogea and its host plants, including the development of proper disease management strategies.
High-throughput phenotyping has a tremendous capacity to advance our understanding of plant biology. Integrating growth parameters with information on a plant's physiology through multispectral imaging can provide a holistic picture of its health status and its responses to environmental stressors. Furthermore, the screening of large-scale populations of genotypes or germplasms, using such platforms, can identify lines with desirable traits to help feed a growing world population in the background of climate change. Here, we present a novel platform, the Multispectral Automated Dynamic Imager (MADI), which combines visible and near-infrared reflectance, thermal imaging, and chlorophyll fluorescence for the dynamic monitoring of growth, leaf temperature, and photosynthetic efficiency. Additionally, we have integrated and validated a fluorescence-based parameter to non-destructively assess chlorophyll content. The utility of the MADI system was demonstrated through four case studies in which lettuce and Arabidopsis plants were exposed to various abiotic stress conditions. We demonstrate that plant compactness is a useful marker for stress responses, including drought, and could serve as a biomarker to study plant hormones. Additionally, we observed the phenomenon of chlorophyll hormesis under salt stress, a rather poorly understood process. In conclusion, the MADI is a multifunctional, adaptable system that can be employed to gain insights into plant stress responses and help to improve agricultural practices. It can be used primarily for rosette-growing species, such as leafy greens, which represent a significant portion of cultivated crops worldwide.
Lettuce, one of the most consumed leafy greens globally, offers significant health benefits due to its high vitamin, mineral, and fiber content. However, Fusarium wilt, a soil-borne fungus, threatens lettuce yields by reducing both quality and quantity. Traditional disease detection methods, such as manual inspection, are time-consuming and inefficient. This study proposes a Unmanned Aerial Vehicle (UAV)-based approach for detecting Fusarium wilt in lettuce using high-resolution Red-Green-Blue (RGB) imagery. (1) a high resolution RGB lettuce dataset captured by drones at approximately 10 m altitude in collaboration with the Yuma Center of Excellence for Desert Agriculture, (2) identification of candidate Fusarium-infected regions by evaluating 300x300 pixel image patches for light tan coloration, followed by the application of a customized Residual Neural Network (ResNet), called LeafyResNet, to confirm Fusarium presence, and (3) a method for quantifying Fusarium infection severity, which was validated against an expert-ground truth. Our approach to detect Fusarium wilt achieves 96.30% accuracy, 94.10% precision, 100% recall, and a 97.10% F1-score, with a 4% false positive rate. Disease severity scores showed an overall accuracy of 86%. We compared the model to state-of-the-art models, including two variants of ResNet (ResNet18 and ResNet34), Inception_v3, and VGG16. LeafyResNet showed superior results compared to available standard models, highlighting the potential of customizing models for agricultural applications. LeafyResNet provides an efficient and scalable solution for Fusarium wilt monitoring for lettuce crops to advance precision agriculture.
Producing food is one of the challenges in space exploration due to limited storage capacity and long travel duration. Extreme environmental conditions such as microgravity, elevated CO2 levels, irregular light exposure, and fluctuating air temperatures pose significant challenges to conventional plant growth and make it susceptible to stress, particularly in root systems, which struggle to absorb water and nutrients efficiently. This study will focus on root phenotyping of the plants (wheat and lettuce) grown in a near-space environment, and the impact of environmental stressors on the plants using image-based technology will be carried out. A specialized growth chamber is designed, incorporating three automated multi-modal imaging systems (MIS): Visible and Near-Infrared (VNIR) wavelength range (400-1000 nm), Micro CT Scan, and RGB cameras used to observe the impact of stress on microgravity on plants. Machine learning and deep learning techniques were also employed to optimize the discriminant classifier within the multi-modal imaging system. Through comparative analysis of these imaging techniques coupled with artificial intelligence techniques, this study aims to deepen our understanding of how microgravity and other space-induced factors affect root systems. This work will also present the challenges and potential faced that can contribute valuable insights for plant growth under space conditions.
Image analysis can be useful for assessing crop health and predicting yield. Instead of expensive equipment, smartphones are considered an accessible and low-cost alternative. The objectives of this study were to evaluate whether fresh weight in green and red lettuce could be predicted by leaf color (intensity of green color measured by RGB) under different fertilizer treatments using RGB imaging from two widely used smartphone models (Samsung Galaxy and Apple iPhone). The two smartphones showed similar longitudinal patterns of RGB data (the intensity and dark green proportion), but the absolute difference in the RGB data was significantly different. Therefore, the averaged results were used for the analyses. Color intensity and dark green proportion were associated with the fresh lettuce weight (p = 0.005, 0.003, 0.014 and p < 0.001, respectively). This study suggests that farmers and practitioners can use these economic devices as a non-destructive method to diagnose and monitor the nutritional status and predict lettuce yield.
Reproduction assets found保存済みの本文根拠を更新済みルールで再検証し、公開資産1件を確認しました。Code · publicThe separated GIF files were uploaded to the image analysis program freely available at http://mkwak.org/imgarea .Open asset ↗lines:321-349Plant phenotyping relevance matchEurope PMC · checked 15 Sept 2026
Published1 May 2025Computers and Electronics in Agriculture.
Accurate measurement of structural phenotypes, such as plant height and canopy width, is crucial for the scientific management of lettuce cultivation in Plant Factories with Artificial Lighting (PFALs). In this study, we developed a multimodal image fusion model using visible images (RGB), depth images (Depth), and infrared images (IR) to extract lettuce phenotypes in PFAL environments. We proposed a Residual Space information enhancement module (DRS) and a fusion feature supplement method with adaptive weight optimization for IR features (IRC) to address the weak spatial perception of traditional RGB-based models and the feature loss of RGB due to illumination disturbance. Three lettuce varieties (Bixiao, Huqian, and Mondai) were selected as experimental subjects to evaluate the robustness of our proposed model. In ablation experiments, the benchmark model improved by DRS increased by 1.6% and 0.9% in terms of MAP0.75 and MAP0.5:0.95, respectively. The benchmark model improved by IRC increased by 0.2%, 0.6%, and 1.2% in terms of MAP0.5, MAP0.75, and MAP0.5:0.95, respectively. Furthermore, MAP0.5, MAP0.75, and MAP0.5:0.95 values increased by 0.3%, 3.3%, and 2.3% when the two modules were combined, respectively. Compared with manually measured plant height and canopy width, the Root Mean Square Error (RMSE) of the average plant height prediction results for the three varieties is 0.74, and the Mean Squared Error (MSE) is 0.55. For canopy width, the RMSE of the model’s prediction results was 0.70, and the MSE was 0.49. In the lighting influence experiment, our method outperformed the unimproved model by approximately 0.3–4% in terms of MAP0.5, MAP0.75, and MAP0.5:0.95 across multiple datasets. Our proposed model effectively addresses lighting disturbance, enhances the robustness of the baseline model against varying lighting conditions, improves spatial perception capability, facilitates the separation of adjacent plant features in the model’s feature extraction stage, enhances the model’s detection ability, and ultimately improves phenotype extraction capability.
Accurate data acquisition of crop morphological parameters is crucial for effective greenhouse management decision-making and remote sensing technologies are increasingly being applied to automate the data collection process. This research utilised an RGB-D based computer vision method to investigate the correlation between the computer vision features and the lettuce morphological parameters, including leaf area, plant height, diameter, and fresh weight. A dataset of lettuce containing over 300 RGB images and depth images of the 3rd Autonomous Greenhouse Challenge was used, and Random Forest, XGBoost and linear regression models were applied in the prediction. The best NRMSE values for diameter, dry matter content, dry weight, fresh weight, height, and leaf area are 0.08, 0.08, 0.07, 0.07, 0.08, and 0.07, which showed a promising accuracy compared to similar studies. This research demonstrates a novel approach to non-destructively estimate greenhouse leafy vegetable morphological parameters.
Nitrogen is a crucial environmental factor influencing lettuce growth, development, and quality formation. This study aimed to determine the relationship between plant growth, nutritional quality formation, and different nitrogen levels of lettuce. A machine learning approach was also applied to data collected from RGB and hyperspectral imaging systems. Traditional methods for nitrogen diagnosis in lettuce, such as laboratory-based analysis of plant samples, are labor-intensive, time-consuming, and lack real-time monitoring capabilities. In contrast, the deep learning models used in this research can make full use of the original data from imaging systems. Nondestructive techniques have the ability to handle complex relationships in the data, enabling more accurate and efficient nitrogen diagnosis. Collected spectral features were combined with chemometrics, and a lettuce nitrogen regression diagnostic model was trained. Furthermore, lettuce crop growth was assessed using a model development of environmental and plant physiological parameters. Additionally, nitrogen fertilization was precisely assessed using developed models. Lettuce cultivation experiments under different nitrogen levels showed the best physiological and biochemical indicators performance when the nitrogen concentration reached 18.75 mmol·L−1. Using machine learning with hyperspectral reflectance in nitrogen diagnostics, random forest showed excellent performance with the highest R2, MSE, and MAE of 0.7012, 8.940, and 2.1859, respectively. ShuffleNet-v2-1.0 obtained a high R2 of 0.9592, MSE of 132.9974, and MAE of 8.1430 regarding transfer learning and hyperspectral images. Applying the transfer learning technique in RGB images exhibited EfficientNet-v2-s, the best model for precise determination of nitrogen diagnostics, with R2 of 0.9859, MSE of 24.0755, and MAE of 2.3433. Current research comprehensively provides both a theoretical basis and practical solutions for precision nitrogen fertilization in lettuce cultivation. Its implications hold significance for the intelligent management of horticultural crop production.
High-throughput phenotyping has a tremendous capacity to advance our understanding of plant biology . Integrating growth parameters with information on a plant's physiology through multispectral imaging can provide a holistic picture of its health status and its responses to environmental stressors. Furthermore, the screening of large-scale populations of genotypes or germplasms , using such platforms, can identify lines with desirable traits to help feed a growing world population in the background of climate change . Here, we present a novel platform, the Multispectral Automated Dynamic Imager (MADI), which combines visible and near-infrared reflectance, thermal imaging , and chlorophyll fluorescence for the dynamic monitoring of growth, leaf temperature, and photosynthetic efficiency . Additionally, we have integrated and validated a fluorescence-based parameter to non-destructively assess chlorophyll content. The utility of the MADI system was demonstrated through four case studies in which lettuce and Arabidopsis plants were exposed to various abiotic stress conditions. We demonstrate that plant compactness is a useful marker for stress responses, including drought, and could serve as a biomarker to study plant hormones. Additionally, we observed the phenomenon of chlorophyll hormesis under salt stress, a rather poorly understood process. In conclusion, the MADI is a multifunctional, adaptable system that can be employed to gain insights into plant stress responses and help to improve agricultural practices. It can be used primarily for rosette-growing species, such as leafy greens, which represent a significant portion of cultivated crops worldwide.
Carbon dots (CDs) have emerged as promising nanomaterials for bioimaging and stress monitoring due to their unique optical and functional properties. CDs were synthesized using citric acid and o -phenylenediamine via microwave-assisted heating, named as CP-CDs. High-resolution transmission electron microscopy observed an average particle size of 3.65 ± 0.40 nm with graphitic cores. Raman spectroscopy and Fourier transform infrared spectroscopy confirmed diverse functional groups. The CDs exhibited excitation-dependent fluorescence with a peak emission at 432 nm, a high quantum yield of 54.91%, and a fluorescence lifetime of 9.50 ± 0.15 ns, making them highly suitable for bioimaging. Confocal microscopy demonstrated tissue-specific localization in lettuce plant cells. In stem cells, CP-CDs predominantly targeted mitochondria, confirmed by a colocalization with Mito-Tracker Red. In contrast, leaf cells showed selective accumulation at the stomatal openings. Under salt- and heat-induced stress, stem cells exhibited an increase in mitochondrial fluorescence, indicating stress-responsive interactions, whereas leaf cells maintained consistent stomatal localization. Further, enhanced fluorescence from chloroplasts under stress conditions suggested synergistic effects with chlorophyll. Also, stress conditions caused CP-CDs to accumulate at the cell boundaries in stem cells, highlighting their sensitivity to stress-induced changes. These findings demonstrate the optical properties, tissue-specific uptake, and organelle-level localization of CP-CDs, underlining their potential for bioimaging, stress detection, and targeted delivery systems in plants.
Lettuce, a vegetable rich in nutritional and medicinal value, is commonly analyzed as a modern industrial crop through macroscopic factors such as temperature, humidity, and light, but its microscopic characteristics are still underexplored. At the microscopic scale, stomatal characteristics are the most indicative of lettuce growth status and serve as crucial pathways for plant gas exchange and carbon-water cycle regulation. Therefore, stomatal research is an important area in crop breeding and stress analysis, and stomatal feature detection is a key step in this field. Current traditional methods for stomatal feature measurement are inefficient, imprecise, and labor-intensive. This study proposes a method for stomatal feature extraction of lettuce leaves based on an improved U-net network to improve measurement efficiency and accuracy. To this end, a dual symmetric path structure was designed, incorporating two independent decoding paths to separately extract global contextual information and local detail features, effectively integrating multi-scale information during the decoding phase via feature concatenation and convolutional fusion modules. To mitigate edge information loss caused by repeated down-sampling in the U-Net network, a hybrid dilated convolution module was incorporated into the encoding phase, with overlapping pooling replacing standard pooling to enhance the network's precision in recognizing small objects. Furthermore, the network incorporates a CBAM attention mechanism module to strengthen its capacity for extracting effective features. To optimize network performance, the NAdam optimization function was employed to speed up convergence and minimize computational resource consumption. The MFe (Measurement Feature) visualization and interaction system developed using OpenCV enables precise measurement of stomatal major and minor axes, area, and density for lettuce leaves. Experimental results indicate that the improved U-Net network achieved enhancements of 3.31 %, 6.55 %, and 4.08 % in IoU, PA, and MPA metrics, respectively. This confirms the effectiveness of the network modifications, offering a valuable reference for microscopic studies of plant stomata.
In agricultural production, lettuce growth, yield, and quality are impacted by nutrient deficiencies caused by both environmental and human factors. Traditional nutrient detection methods face challenges such as long processing times, potential sample damage, and low automation, limiting their effectiveness in diagnosing and managing crop nutrition. To address these issues, this study developed a lettuce nutrient deficiency detection system using multi-dimensional image analysis and Field-Programmable Gate Arrays (FPGA). The system first applied a dynamic window histogram median filtering algorithm to denoise captured lettuce images. An adaptive algorithm integrating global and local contrast enhancement was then used to improve image detail and contrast. Additionally, a multi-dimensional image analysis algorithm combining threshold segmentation, improved Canny edge detection, and gradient-guided adaptive threshold segmentation enabled precise segmentation of healthy and nutrient-deficient tissues. The system quantitatively assessed nutrient deficiency by analyzing the proportion of nutrient-deficient tissue in the images. Experimental results showed that the system achieved an average precision of 0.944, a recall rate of 0.943, and an F1 score of 0.943 across different lettuce growth stages, demonstrating significant improvements in automation, accuracy, and detection efficiency while minimizing sample interference. This provides a reliable method for the rapid diagnosis of nutrient deficiencies in lettuce.
Reproduction assets foundThe paper's Data Availability Statement explicitly states that the original data, implementation code, and sample data are openly available on the authors' GitHub (https://github.com/lvss88), which matches an allowed URL. This qualifies as a paper-specific public asset covering the lettuce nutrient-deficiency image-d分析Code · publicData Availability Statement: The original data, including implementation code and sample data, pre-
sented in the study are openly available at https://github.com/lvss88 (accessed on 23 January 2025).Open asset ↗lvss88pdf-page:24 lines:1-59Plant phenotyping relevance matchEurope PMC · OpenAlex · checked 13 Sept 2026
Accurate and nondestructive estimation of plant biomass is crucial for optimizing plant productivity, but existing methods are often expensive and require complex experimental setups. To address this challenge, we developed an automated system for estimating plant root and shoot biomass over the plant's lifecycle in hydroponic systems. This system employs a robotic arm and turntable to capture 40 images at equidistant angles around a hydroponically grown lettuce plant. These images are then processed into silhouettes and used in voxel-based volumetric 3D reconstruction to produce detailed 3D models. We utilize a space carving method along with a raytracing-based optical correction technique to create high-accuracy reconstructions. Analysis of these models demonstrates that our system accurately reconstructs the plant root structure and provides precise measurements of root volume, which can be calibrated to indicate biomass.
Accurate identification of high-quality seeds is crucial for maintaining superior crop traits. Lettuce is widely consumed vegetable with diverse varieties, however, the traditionally identification methods are both time-consuming and labor-intensive. This study explores feasibility of rapid, non-destructive identification of different lettuce varieties using multispectral imaging combined with machine learning. We firstly collected seed morphological and spectral data from 15 lettuce varieties using multispectral imaging. Then we applied Support Vector Machine (SVM), Random Forest (RF), and Back-Propagation Neural Network (BP), Linear Discriminant Analysis (LDA) for variety identification. The results demonstrated that multispectral imaging combined with machine learning models, effectively distinguished different lettuce seed varieties. The LDA model based on morphological and spectral fusion feature data performed best, and the average classification accuracy was 92.7 %. In the batch validation, the LDA model achieved an accuracy of 93.2 %.This method reduces cost and improves efficiency, showing great potential for seed identification in other crops.
Daniil S. Ulyanov · Alana A. Ulyanova · Dmitry Litvinov · А.А. Кочешкова · Aleksandra Yu. Kroupina · Nadejda M. Syedina · V.S. Voronezhskaya · Andrey V. Vasilyev · Gennady I. Karlov · М.Г. Дивашук
Digital phenotyping is a fast-growing area of hardware and software research and development. Phenotypic studies usually require determining whether there is a difference in some trait between plants with different genotypes or under different conditions. We developed StatFaRmer, a user-friendly tool tailored for analyzing time series of plant phenotypic parameters, ensuring seamless integration with common tasks in phenotypic studies. For maximum versatility across phenotypic methods and platforms, it uses data in the form of a set of spreadsheets (XLSX and CSV files). StatFaRmer is designed to handle measurements that have variation in timestamps between plants and the presence of outliers, which is common in digital phenotyping. Data preparation is automated and well-documented, leading to customizable ANOVA tests that include diagnostics and significance estimation for effects between user-defined groups. Users can download the results from each stage and reproduce their analysis. It was tested and shown to work reliably for large datasets across various experimental designs with a wide range of plants, including bread wheat (Triticum aestivum), durum wheat (Triticum durum), and triticale (× Triticosecale); sugar beet (Beta vulgaris), cocklebur (Xanthium strumarium) and lettuce (Lactuca sativa), corn (Zea mays) and sunflower (Helianthus annuus), and soybean (Glycine max). StatFaRmer is created as an open-source Shiny dashboard, and simple instructions on installation and operation on Windows and Linux are provided.
Reproduction assets foundThe paper's authors publicly release StatFaRmer, an open-source R Shiny dashboard for phenotyping data analysis, via GitHub with installation instructions and a sample phenotypic dataset, and host a live deployment on shinyapps.io.Code · publicThe resulting tool can be accessed at 9 https://github.com/Stathmin/StatFaRmer ), with the instructions on installation and the sample dataset provided.Open asset ↗Stathmin/StatFaRmerlines:521-528Dataset · publicA sample dataset of different plant species (bread wheat ( Triticum aestivum ), durum wheat ( Triticum durum ), and triticale (× Triticosecale )), cultivars (35 variants) and plant genotypes (allelic state of 3 genes), with different treatments (3 variants), and the time series of morphological and spectral parameters of these plants is loaded in this tool as an example and available on GitHub.Open asset ↗lines:340-350Plant phenotyping relevance matchOpenAlex · checked 15 Sept 2026
Multispectral imaging plays a key role in crop monitoring. A major challenge, however, is spectral band misalignment, which can hinder accurate plant health assessment by distorting the calculation of vegetation indices. This study presents a novel approach for short-range calibration of a multispectral camera, utilizing stereo vision for precise geometric correction of acquired images. By using multispectral camera lenses as binocular pairs, the sensor acquisition distance was estimated, and an alignment model was developed for distances ranging from 500 mm to 1500 mm. The approach relied on selecting the red band image as a reference, while the remaining bands were treated as moving images. The stereo camera calibration algorithm estimated the target distance, enabling the correction of band misalignment through previously developed models. The alignment models were applied to assess the health status of baby leaf crops (Lactuca sativa cv. Maverik) by analyzing spectral indices correlated with chlorophyll content. The results showed that the stereo vision approach used for distance estimation achieved high accuracy, with average reprojection errors of approximately 0.013 pixels (4.485 × 10−5 mm). Additionally, the proposed linear model was able to explain reasonably the effect of distance on alignment offsets. The overall performance of the proposed experimental alignment models was satisfactory, with offset errors on the bands less than 3 pixels. Despite the results being not yet sufficiently robust for a fully predictive model of chlorophyll content in plants, the analysis of vegetation indices demonstrated a clear distinction between healthy and unhealthy plants.
Lettuce is one of the major raw vegetables in the world, with diverse species and large differences in morphological structures. Achieving automated, high-throughput acquisition and intelligent analysis of 3D lettuce phenotypes using advanced phenotyping techniques and equipment is of great significance. Based on the high-throughput phenotyping platform MVS-PhenoV2 installed in a plant imaging room, this study constructed a method for automated analysis of 3D phenotypes of lettuce around the needs of lettuce DUS (distinctiveness, uniformity, and stability) testing and feature digitisation. Aiming at the characteristics of lettuce leaves which are mostly curved, the point cloud segmentation model SoftGroup was improved, which can realise lettuce single plant segmentation and leaf segmentation with high accuracy. Additionally based on lettuce 3D point clouds, plant orientation correction algorithm, leaf hole completion algorithm, leaf vein extraction algorithm, and leaf margin extraction algorithm were proposed. Finally, a pipelined automated analysis software tool LettuceP3D was developed for automated analysis of lettuce 3D phenotypes, which can automatically analyse 16 phenotypic indicators related to lettuce plant (e.g. plant height, plant width, and compactness) and leaf (e.g. leaf length, leaf margin perimeter, and leaf margin undulation) phenotypic characteristics. The study was validated on seven types of lettuce: Butterhead, Crisphead, Looseleaf, Oakleaf, Romaines, Stem, and Wild Relatives. Results show that the mIoU for plant and pot semantic segmentation reaches 97.2%, and the AP for leaf instance segmentation reaches 86.7%. Through comparison with measured values, the average R2 of the algorithm exceeds 0.95. The software operates without manual interaction, processing single plant data in approximately 2s, which demonstrates a high processing efficiency. This phenotype analysis method proposed in this study is applicable for quantifying the morphological characteristics of lettuce in seven types, providing quantitative indicator data support for lettuce DUS testing, variety identification, and multi-omics studies.
In modern agricultural technology, the use of computer vision and deep learning methods for high-throughput phenotypic analysis of crops has become a key trend in improving agricultural production efficiency and accuracy. Especially in the area of instance segmentation, precise and efficient field crop image segmentation allows for faster and more accurate acquisition of crop field phenotypic traits, which is of significant value for disease identification, growth monitoring, and yield prediction, among other aspects. To this end, we propose a precise and efficient instance segmentation network named YOMASK. This network integrates various advanced technologies, including feature extraction, feature fusion, and attention mechanisms, optimizing the accuracy of the detection and segmentation process. Moreover, the role of each module in task execution is verified through visualization methods, enhancing the transparency and interpretability of the model’s internal decision-making process. Tested on a high-throughput phenotyping platform (HTPP) for the instance segmentation task of lettuce, YOMASK exhibited outstanding performance, achieving a detection accuracy of 94.52 % and a segmentation accuracy of 95.41 %, with a model size of 19.9 MB and an inference speed of 103.9FPS. Compared to existing instance segmentation models such as Mask RCNN, SOLOv2, and YOLACT, YOMASK has shown significant improvements in both accuracy and efficiency, effectively detecting each lettuce instance in the image and generating high-quality segmentation masks for them. This research is of significant importance in the field of precision agriculture, especially in high-throughput phenotyping analysis and crop health monitoring.
Abstract Monitoring plant growth is crucial for effective crop management, and using color and depth (RGBD) cameras to model lettuce has emerged as one of the most convenient and non-invasive methods. In recent years, deep learning techniques, particularly neural networks, have become popular for estimating lettuce fresh weight. However, these models are typically specific to particular datasets, lack domain adaptation, and are often limited by the availability of open-access datasets. In this study, we propose a method based on plant geometric features for estimating the rosette structure and volume of lettuce. This new approach was compared to existing methods that reconstruct surfaces from point clouds, such as Ball Pivoting and Alpha Shapes. The proposed method creates a tight hull around the plant's point cloud, preserving high detail of the rosette structure while filling in surface holes in areas not visible to 3D cameras. Using a linear regression model, we estimated fresh weight for this dataset, achieving a root mean square error (RMSE) of 18.2 g when using only the estimated plant volume, and 17.3 g when both volume and geometric features were included. Additionally, we introduced new geometric features that characterize leaf density, which could be useful for breeding applications. A dataset of 402 point clouds of lettuce plants, captured before harvest, was compiled using one top-down and three side-view 3D cameras.
Reproduction assets foundThe paper's own lettuce 3D point cloud dataset (Pii, 402 point clouds with fresh weight references) is deposited on Zenodo, and the vacuum-package surface reconstruction code plus data processing scripts are publicly available on the authors' GitHub repository. Both are paper-specific, public, and actionable.Dataset · publicData used in this study and developed models are available on Zenodo storage service https://zenodo.org/records/8410252 .Open asset ↗Zenodo · 8410252lines:158-220Code · publicThe code used at this study is available at https://github.com/VicB18/LettuceFW (accessed on 1 November 2024).Open asset ↗GitHub · VicB18/LettuceFWlines:158-220Code · publicThe code for the vacuum package method, along with the data processing scripts used in this study, are available in the Supplementary Information.Open asset ↗lines:98-114Plant phenotyping relevance matchOpenAlex · checked 15 Sept 2026
Lettuce is one of the major raw vegetables in the world, with diverse species and large differences in morphological structures. Achieving automated, high-throughput acquisition and intelligent analysis of 3D lettuce phenotypes using advanced phenotyping techniques and equipment is of great significance. Based on the high-throughput phenotyping platform MVS-PhenoV2 installed in a plant imaging room, this study constructed a method for automated analysis of 3D phenotypes of lettuce around the needs of lettuce DUS (distinctiveness, uniformity, and stability) testing and feature digitisation. Aiming at the characteristics of lettuce leaves which are mostly curved, the point cloud segmentation model SoftGroup was improved, which can realise lettuce single plant segmentation and leaf segmentation with high accuracy. Additionally based on lettuce 3D point clouds, plant orientation correction algorithm, leaf hole completion algorithm, leaf vein extraction algorithm, and leaf margin extraction algorithm were proposed. Finally, a pipelined automated analysis software tool LettuceP3D was developed for automated analysis of lettuce 3D phenotypes, which can automatically analyse 16 phenotypic indicators related to lettuce plant (e.g. plant height, plant width, and compactness) and leaf (e.g. leaf length, leaf margin perimeter, and leaf margin undulation) phenotypic characteristics. The study was validated on seven types of lettuce: Butterhead, Crisphead, Looseleaf, Oakleaf, Romaines, Stem, and Wild Relatives. Results show that the mIoU for plant and pot semantic segmentation reaches 97.2%, and the AP for leaf instance segmentation reaches 86.7%. Through comparison with measured values, the average R 2 of the algorithm exceeds 0.95. The software operates without manual interaction, processing single plant data in approximately 2s, which demonstrates a high processing efficiency. This phenotype analysis method proposed in this study is applicable for quantifying the morphological characteristics of lettuce in seven types, providing quantitative indicator data support for lettuce DUS testing, variety identification, and multi-omics studies.
Monitoring the growth dynamics of plants in three-dimensional (3D) space is one of the most fundamental data acquisition requirements for plant breeding and cultivation. The rapid development of high-throughput plant phenotyping platforms (HTPPP) makes it possible to obtain big data in plant phenomics. However, how to extract phenotypes from the raw phenotyping data to obtain the agronomic indicators demanded by agronomists has become an urgent issue. In this study, time-series point clouds of potted lettuce plants were generated via multi-view stereo (MVS) method using top-view Red, Green, Blue (RGB) images acquired by a rail-driven HTPPP in a greenhouse. A time-series point cloud registration method was proposed by extracting pots as features, and daily population-individual plant point cloud segmentation was achieved based on the registration information and contrasted with two other different segmentation methods. Then vegetation and pot was segmented using the random forest (RF). Finally, the phenotypes including plant height, crown width, and convex hull volume of each plant were extracted. The results show that the average mean intersection over union (mIoU), mean precision (mPᵣ), mean recall (mRₑ), and mean F1-score (mF₁) of the population-individual plant segmentation were 71.86%, 97.38%, 86.08%, and 91.02%, respectively. The vegetation-pot point cloud segmentation achieved an accuracy of 98.81%. The averaged coefficient of determination (R²) for the extracted plant height and crown width were 0.79 and 0.60, respectively, with the averaged root mean square error (RMSE) being 0.05 m and 0.03 m, respectively. The accuracy of plant height was significantly higher than that of PlantEye. The extracted phenotypes can be used to quantitatively differentiate the growth dynamics of different sub-populations of lettuce plants. This study presents an automated solution for extracting time-series 3D phenotypes under HTPPP in a greenhouse. It provides crucial technological support for efficient phenotype acquisition in plant breeding and cultivation.
With the application of facility-based agriculture, robotics, and other technologies in agricultural production, the accurate recognition of individual crop phenological stages has become crucial for precision agriculture. However, the slightly noticeable visual disparities between neighboring phenological stages of crops pose a challenge for fine-grained phenological classification based on image features. Phenotypic features of plants, providing a wealth of information about crop-growth patterns, exhibit remarkable similarity within the same phenological stage. This paper presents a framework named PhenologyNet, which leverages the significant role of phenotypic features for fine-grained crop-phenology classification. A novel phenology-classification model was developed by leveraging local crop phenotypic features, assigning dynamic weights based on a phenology-specific matrix. The model calculates overall similarity from nine patch pair similarities, capturing the nuanced importance of local features during different phenological stages. This was combined with the phenology classification based on convolutional neural networks (CNNs) using a late fusion approach, achieving precise fine-grained recognition of crop phenological stages. PhenologyNet was trained and tested on a lettuce-phenology dataset classified according to 28 “Biologische Bundesanstalt, Bundessortenamt und Chemische Industrie” (BBCH) phenological stages, achieving an accuracy of 92.79%. Comparative experiments with other methodologies excluding phenotypic similarity showed that the incorporation of phenotypic similarity notably enhanced the efficiency of the fine-grained crop-phenology classification. PhenologyNet’s performance further improved when the number of BBCH phenological stages decreased, suggesting the framework’s commendable capabilities in tackling the issue of crop-phenology classification in a way that meets precision agriculture’s requirements.
Crop monitoring can promptly reflect the growth status of crops. However, conventional methods of growth monitoring, although simple and direct, have limitations such as destructive sampling, reliance on human experience, and slow detection speed. This study estimated the fresh weight of lettuce (Lactuca sativa L.) in a plant factory with artificial light based on three-dimensional (3D) reconstruction technology. Data from different growth stages of lettuce were collected as the training dataset, while data from different plant forms of lettuce were used as the validation dataset. The partial least squares regression (PLSR) method was utilized for modeling, and K-fold cross-validation was performed to evaluate the model. The testing dataset of this model achieved a coefficient of determination (R2) of 0.9693, with root mean square error (RMSE) and mean absolute error (MAE) values of 3.3599 and 2.5232, respectively. Based on the performance of the validation set, an adaptation was made to develop a fresh weight estimation model for lettuce under far-red light conditions. To simplify the estimation model, reduce estimation costs, enhance estimation efficiency, and improve the lettuce growth monitoring method in plant factories, the plant height and canopy width data of lettuce were extracted to estimate the fresh weight of lettuce in addition. The testing dataset of the new model achieved an R2 value of 0.8970, with RMSE and MAE values of 3.1206 and 2.4576.
Semantic segmentation methods have played an important role in a wide range of applications, as they contribute to more accurate phenotypic information extraction in the field of plant phenotype. However, the high annotation cost of semantic segmentation datasets remains a major challenge, and most of them are constructed and validated on training and testing datasets with similar scales. Most studies overlook its effectiveness on multi-scale datasets, especially on low resolution datasets. Although some semantic segmentation methods extract and learn multi-scale features from datasets through methods such as multi-scale feature fusion modules and attention mechanisms, the model's scale down compatibility, i.e. the segmentation reliability of the model on low resolution datasets, has not yet been verified. To address this challenge, this study proposes for the first time a new approach to plant object oriented semantic segmentation, which involves modeling individual target datasets and validating group target datasets. This modeling approach can significantly reduce the annotation cost of datasets to some extent. On this basis, we propose a multi-scale feature fusion module (MSFAF-M) for multi-level feature relationship exploration and a multi receptive field feature fusion module (MRFFF-S) for single-layer feature relationship exploration. By applying MSFAF-M and MRFFF-S to U2Net, an upgraded semantic segmentation method MRSU2Net is proposed, which can fully extract global and local feature information of target objects at multiple scales, and improve the segmentation reliability of semantic segmentation models based on individual target datasets on multi-scale group target datasets. Due to the fact that the construction approach of the semantic segmentation model proposed in this study is different from traditional semantic segmentation methods, we validated the scale down compatibility of MRSU2Net on the target dataset of lettuce populations collected at the seedling stage. When MRSU2Net is applied to group target images with the same resolution (2992 × 2992), the MIoU is 0.9719 and the inference-time is 0.3550. When MRSU2Net is applied to group target images of the same input size (224 × 224), the MIoU can reach 0.7346 and the inference time is 0.0219. The results demonstrate that the segmentation performance of the MRSU2Net constructed in this study is significantly superior to other classic semantic segmentation methods in low resolution images.
To the optimal time to conduct farming operations in the traditional agricultural production process mainly depends on human observation and planting experience, which is time-consuming and laborious, and makes it easy to miss the best agricultural operation opportunities. In this study, our main objective is to accurately detect the key growth stages of lettuce to guide the timely implementation of corresponding agricultural operations. Firstly, the dataset was collected for the growth stage with important agricultural operations in the growth process of multi-variety lettuce, to lay the data foundation for the construction of the model. Secondly, considering the difference in plant growth, we compared many methods and selected the optimal modeling method YOLOXs to identify the key growth stages of multi-variety lettuce (mAP = 98.75 %). Finally, to ensure the applicability of the detection model in complex agricultural scenes, we tried to improve the effect of YOLOXs by three attention mechanisms and one multi-scale feature fusion method, and proposed a new method CBAM + ASFF-YOLOXs (mAP = 99.04 %). The results showed that this method is expected to replace human eye observation and experience in planting, to provide accurate technical feedback on relevant agricultural operation time, and to provide technical support for the unmanned operation of agriculture. At the same time, the limitations, challenges, and prospects of this method are discussed.
Monitoring plant growth is crucial for cultivation management. Agronomists can assess the health status of lettuce seedlings based on monitoring results to implement relevant management measures for improving the quality and yield of lettuce seedlings. This study developed a non-destructive, high-throughput growth monitoring method suitable for large-scale assessment of lettuce seedling quality in nurseries. The method utilizes a plant high-throughput phenotyping platform to acquire 10-day time-series imagery data. An Mask2Former network model enhanced by multidimensional collaborative attention mechanism, combined with sliding window and morphological operations, achieves precise recognition and localization of seedling trays, varieties, and individual seedling plants in a progressive manner. Based on individual seedling localization and segmentation results, the method estimates emergence numbers and rates for each variety, and further achieves instance segmentation and counting of individual seedling leaves, innovatively constructing leaf segmentation results of different varieties across the entire seedling tray. Applied to time-series images, the method automatically monitored seedling emergence changes and growth trends for 1,086 lettuce varieties. In monitoring these varieties, the method achieved a coefficient of determination (R²) of 0.96 for emergence number estimation. The extraction of all six key phenotypic parameters demonstrated exceptionally high correlations: projected area, projected perimeter, convex hull area, and convex hull perimeter all showed R² above 0.99, while leaf compactness R² was 0.9698, and leaf count R² was 0.91. Results demonstrate that this high-throughput, reliable method can effectively monitor the growth status of large-scale lettuce seedlings and provide technical support for lettuce nursery quality assessment.
Agriculture is the backbone of the country’s economy. People depend on agriculture for food and exporting to generate income. However, agriculture faces various diseases that affect the quantity and quality of vegetables. Therefore, it is important to propose a model for detecting vegetable diseases. This study proposed a sustainable smart system for vegetable disease detection and classification. This system detects early vegetable diseases in common vegetables such as tomato, potato, lettuce, and cucumber. The study employed deep learning (DL) models to detect and classify vegetable diseases. Convolutional neural networks (CNN) are a type of DL model used for image classification. This study utilizes CNN and other extensions, such as VGG16 and MobileNet, for plant image classification. Three DL models were trained on four datasets for tomato disease classification, potato disease classification, lettuce disease classification, and cucumber disease classification. The results show that the three models achieved 84.49% accuracy on the tomato disease dataset, 97.65% accuracy on the cucumber disease dataset, 97% accuracy on the potato disease dataset, and 99.9% accuracy on the lettuce disease dataset. The proposed system can assist farmers in the early detection of vegetable diseases before they spread, and it can enhance agriculture by improving both the quality and quantity of products.
Effective lettuce cultivation requires precise monitoring of growth characteristics, quality assessment, and optimal harvest timing. In a recent study, a deep learning model based on multimodal data fusion was developed to estimate lettuce phenotypic traits accurately. A dual-modal network combining RGB and depth images was designed using an open lettuce dataset. The network incorporated both a feature correction module and a feature fusion module, significantly enhancing the performance in object detection, segmentation, and trait estimation. The model demonstrated high accuracy in estimating key traits, including fresh weight (fw), dry weight (dw), plant height (h), canopy diameter (d), and leaf area (la), achieving an R2 of 0.9732 for fresh weight. Robustness and accuracy were further validated through 5-fold cross-validation, offering a promising approach for future crop phenotyping.
Reproduction assets foundThe paper's RGB-D lettuce images and trait measurements come from the publicly available Third Autonomous Greenhouse Challenge dataset deposited at 4TU.ResearchData, with an explicit availability statement and URL matching an allowed URL. No author analysis code or trained model is disclosed.Dataset · publicThis study used the Third Autonomous Greenhouse Challenge: Online Challenge Lettuce Images dataset publicly available at 4TU.ResearchData [ 36 ].Open asset ↗4TU.ResearchDatalines:819-832Code / dataset availability confirmedOpenAlex · Europe PMC · bioRxiv · checked 13 Sept 2026
Rens Dijkhuizen · Abraham L. van Eijnatten · Sarah L. Mehrem · Esther S. van den Bergh · Jelmer van Lieshout · Kiki Spaninks · Steven Kaandorp · Remko Offringa · Marcel Proveniers · Guido Van den Ackerveken · Basten L. Snoek
Abstract In recent years, the automation of genotyping has significantly enhanced the efficiency of genome-wide association studies. As a result, phenotyping rather than genotyping is now the rate-limiting step, especially in field experiments. For this reason, there is a strong need to further automate in-field phenotyping. Here we present a GWAS study on 194 field-grown accessions of lettuce ( Lactuca sativa ). These accessions were non-destructively phenotyped at two time points 15 days apart using an unmanned aerial vehicle. Our high throughput phenotyping approach integrates an RGB camera, a multispectral camera to measure the reflectance at 5 wavelengths (blue, green, red, red edge, near-infrared), and precise height estimation. We used the mean and other descriptives such as median, quantiles, minimum and maximum to quantify different aspects of color and height variation in lettuce from the drone images. Using this approach, we confirm several previously described QTLs, now in populations grown under field conditions, and identify several new QTLs for plant-height and color.
Reproduction assets foundThe paper explicitly states that analysis scripts are publicly available on GitHub (SnoekLab/Dijkhuizen_etal_2025_Drone) and that extended data (raw data, intermediate steps, figure data, weather data) is deposited at the Utrecht University repository DOI 10.24416/UU01-S5FCM9. Both are paper-specific, public, and verbiCode · publicScripts used for this study are available on github:
https://github.com/SnoekLab/Dijkhuizen_etal_2025_Drone.Open asset ↗SnoekLab/Dijkhuizen_etal_2025_Dronepdf-page:8 lines:1-120Dataset · publicExtended data available on https://doi.org/10.24416/UU01-S5FCM9. This includes all raw
data to reproduce results, all intermittent steps, the data required to generate all figures and
the weather data.Open asset ↗10.24416/UU01-S5FCM9pdf-page:8 lines:1-120Plant phenotyping relevance matchEurope PMC · checked 7 Sept 2026
Crop growth information is collected through destructive investigation, which inevitably causes discontinuity of the target. Real-time monitoring and estimation of the same target crops can lead to dynamic feedback control, considering immediate crop growth. Images are high-dimensional data containing crop growth and developmental stages and image collection is non-destructive. We propose a non-destructive growth prediction method that uses low-cost RGB images and computer vision. In this study, two methodologies were selected and verified: an image-to-growth model with crop images and a growth simulation model with estimated crop growth. The best models for each case were the vision transformer (ViT) and one-dimensional convolutional neural network (1D ConvNet). For shoot fresh weight, shoot dry weight, and leaf area of lettuce, ViT showed R 2 values of 0.89, 0.93, and 0.78, respectively, whereas 1D ConvNet showed 0.96, 0.94, and 0.95, respectively. These accuracies indicated that RGB images and deep neural networks can non-destructively interpret the interaction between crops and the environment. Ultimately, growers can enhance resource use efficiency by adapting real-time monitoring and prediction to feedback environmental controls to yield high-quality crops.
Automatically identifying key physiological factors in plants, such as leaf relative humidity (LRH), chlorophyll content (Chl), and nitrogen levels (N), is vital for effective aeroponic management and improving growth, yield, quality, and sustainability. Meta-learning (MetaL) solutions utilize data fusion and intelligent processing, ensuring fast and consistent outcomes. This paper aims to develop a novel MetaL framework that leverages multimodal data sources—including spectral, thermal, and IoT environmental data—to enable real-time, non-invasive identification of LRH, Chl, and N content in aeroponically grown lettuce. The research examined various spectral reflectance indices (SRIs) and thermal indicators from plant characteristics. Model-based feature selection was implemented using back-propagation neural networks (BPNN), decision trees (DT), and gradient boosting machines (GBM) to identify key attributes and optimize hyperparameters. The experimental findings indicated that deploying GBM-based top variables as the foundational model, combined with BPNN as the meta-model, significantly improved the accuracy of analyzing the assigned factors. The prediction scores (R²) for LRH, Chl, and N increased to 0.875 (RMSE=0.879), 0.886 (RMSE=0.694), and 0.930 (RMSE=0.184), respectively, compared to applying BPNN-based features alone as a standalone model. Overall, the designed methodology contributes to more accurate predictions of plant physiological states, enabling proactive steps toward sustainable aeroponic agriculture.
Abstract In the realm of global food security, plants serve as the primary source of sustenance. However, plant diseases pose a significant threat to this security. The process of diagnosing these diseases forms the bedrock of disease control efforts. The precision and expediency of these diagnoses wield substantial influence over disease management and the consequent reduction of economic losses. Conversely, incorrect diagnoses can render interventions ineffective, leading to agricultural crop deterioration and compounding economic hardships for both farmers and their respective nations. This research endeavors to diagnose the prevalent crops in Jordan, as identified by the Jordanian Department of Statistics for the year 2019. These crops encompass four key agricultural varieties: cucumbers, tomatoes, lettuce, and cabbage. To facilitate this, a novel dataset known as "Jordan 22" was meticulously curated. Jordan 22 was painstakingly compiled through the collection of images featuring both diseased and healthy plants, captured within the confines of Jordanian farms. These images underwent meticulous classification by a panel of three agricultural specialists, well-versed in plant disease identification and prevention. The Jordan 22 dataset comprises a substantial size, amounting to 3210 images. Following the compilation of this dataset, a series of preprocessing steps were executed. These encompassed the standardization of image backgrounds and the uniformization of image dimensions. Furthermore, image augmentation techniques were applied to the dataset to expand its diversity. Subsequently, a deep learning model, the Convolutional Neural Network (CNN), was meticulously trained on the augmented dataset. The results yielded by the CNN were nothing short of remarkable, with a test accuracy rate reaching an impressive 0.9712. Optimal performance was observed when images were resized to 256x256 dimensions, and max pooling was employed in lieu of average pooling within the pooling layer. Furthermore, the initial convolutional layer was set at a size of 32, with subsequent convolutional layers standardized at 128 in size. In conclusion, this research represents a pivotal step towards enhancing plant disease diagnosis and, by extension, global food security. Through the creation of the Jordan 22 dataset and the meticulous training of a CNN model, we have achieved substantial accuracy in disease detection, paving the way for more effective disease management strategies in agriculture.
Reproduction assets foundThe paper's Jordan22 plant disease image dataset (2310 RGB leaf images of cucumber, tomato, cabbage, and lettuce collected in Jordan and expert-classified) is explicitly stated as openly available on the authors' public GitHub repository. No separate analysis code or trained model checkpoint is explicitly deposited.Dataset · publicThe data that support the findings of this study are openly available in [Jordan22_Dataset] at
[https://github.com/shahd1995913/Jordan22_Dataset], reference number [17].Open asset ↗Jordan22_Datasetpdf-page:25 lines:1-40Plant phenotyping relevance matchEurope PMC · checked 15 Sept 2026
Published1 Oct 2024Computers and Electronics in Agriculture.
Fresh weight is a crucial indicator for assessing crop growth in plant factory. To date, the majority of non-destructive techniques employed for estimating crop fresh weight rely on the top view images. Nevertheless, these approaches are limited in performance due to the constrained and non-open environment of plant factories. In this paper, we propose a novel position-guided network (PosNet) to estimate the fresh weight of crop using oblique view images. Precisely, we first build the crop (i.e., lettuce) dataset by positioning the camera vertically in an oblique angle, with the Mask R-CNN framework trained for individual lettuces segmentation from the background. The segmented lettuce images are then fed into the PosNet for network training and testing purpose. By integrating the shallow feature extraction module and the position information extraction module, the proposed PosNet attains superior performance on assessing the lettuce fresh weight from oblique view images against other models. We further conducted the ablation studies and generalization testing to verify the efficacy and robustness of the proposed network model. Moreover, by comparing three variations (i.e., position, growth stage and posture orientation) of the lettuce separately and the sensitivity analysis of oblique shooting angles, our method demonstrates plausible adaptability for lettuce fresh weight estimation. Taking the accuracy, robustness, generalization capability, and adaptability into account, the integration of PosNet with oblique images not only enjoys great potentials in assessing the fresh weight, but also provides a practical support for agronomic management of crops cultivated in plant factories.
Abstract Microscopic imaging for studying plant-pathogen interactions is limited by its reliance on invasive histological techniques, like clearing and staining, or, for in vivo imaging, on complicated generation of transgenic pathogens. We present real-time 3D in vivo visualization of pathogen dynamics with label-free optical coherence tomography. Based on intrinsic signal fluctuations as tissue contrast we image filamentous pathogens and a nematode in vivo in 3D in plant tissue. We analyze 3D images of lettuce downy mildew infection ( Bremia lactucae ) to obtain hyphal volume and length in three different lettuce genotypes with different resistance levels showing the ability for precise (micro) phenotyping and quantification of the infection level. In addition, we demonstrate in vivo longitudinal imaging of the growth of individual pathogen (sub)structures with functional contrast on the pathogen micro-activity revealing pathogen vitality thereby opening a window on the underlying molecular processes.
Reproduction assets foundThe authors explicitly deposited supporting code for dynamic OCT processing, segmentation, and data analysis, together with a representative selection of the dynamic OCT volumes (the paper's plant-pathogen phenotyping data), in a freely-accessible Zenodo repository (10.5281/zenodo.11428245). This is a paper-specific,公开Dataset · publicA representative selection of the data, all the dynamic OCT volumes, and supporting code for data processing and plotting have been uploaded to a freely-accessible Zenodo repository 33 . [10.5281/zenodo.11428245].Zenodo · 10.5281/zenodo.11428245lines:133-155Plant phenotyping relevance matchEurope PMC · checked 15 Sept 2026
Published19 Sept 2024Analytical methods : advancing methods and applicationsCited by 3 · OpenAlex ↗
As an important signaling molecule, carbon monoxide (CO) plays an important role in plant growth and development including affecting stomatal movement, stress response and root development. Thus, it is necessary to develop fluorescent probes that can be used to detect CO in live plant tissues and further enable a deep-understanding of its biological function, mechanism and metabolism. In this paper, a novel and sensitive fluorescent probe based on Cu 2+ modulated polydihydroxyphenylalanine nanoparticles (PDOAs) has been developed for the detection of CO. The fluorescence of PDOAs can be effectively quenched by Cu 2+ through the multi-coordination interaction. In the presence of CO, Cu 2+ can be effectively reduced to Cu + , which resulted in the release of free PDOAs and the Cu 2+ -quenched bright green fluorescence was restored obviously. Through this ingenious strategy, the abiotic CO can be accurately detected and identified with high selectivity, rapid response time within 5 min and an ultralow detection limit of 72.4 nM. Due to the admirable biocompatibility, the nano-material based probe has been successfully applied for in vivo imaging CO in the root tip and leave tissues of lettuce. To the best of our knowledge, this is the first example of a fluorescent probe-based methodology for the sensitive tracking of CO in plant tissues.
Modern agriculture is characterized by the use of smart technology and precision agriculture to monitor crops in real time. The technologies enhance total yields by identifying requirements based on environmental conditions. Plant phenotyping is used in solving problems of basic science and allows scientists to characterize crops and select the best genotypes for breeding, hence eliminating manual and laborious methods. Additionally, plant phenotyping is useful in solving problems such as identifying subtle differences or complex quantitative trait locus (QTL) mapping which are impossible to solve using conventional methods. This review article examines the latest developments in image analysis for plant phenotyping using AI, 2D, and 3D image reconstruction techniques by limiting literature from 2020. The article collects data from 84 current studies and showcases novel applications of plant phenotyping in image analysis using various technologies. AI algorithms are showcased in predicting issues expected during the growth cycles of lettuce plants, predicting yields of soybeans in different climates and growth conditions, and identifying high-yielding genotypes to improve yields. The use of high throughput analysis techniques also facilitates monitoring crop canopies for different genotypes, root phenotyping, and late-time harvesting of crops and weeds. The high throughput image analysis methods are also combined with AI to guide phenotyping applications, leading to higher accuracy than cases that consider either method. Finally, 3D reconstruction and a combination with AI are showcased to undertake different operations in applications involving automated robotic harvesting. Future research directions are showcased where the uptake of smartphone-based AI phenotyping and the use of time series and ML methods are recommended.
Precision agriculture (PA) technologies combined with remote sensors, GPS, and GIS are transforming the agricultural industry while promoting sustainable farming practices with the ability to optimize resource utilization and minimize environmental impact. However, their implementation faces challenges such as high computational costs, complexity, low image resolution, and limited GPS accuracy. These issues hinder timely delivery of prescription maps and impede farmers’ ability to make effective, on-the-spot decisions regarding farm management, especially in stress-sensitive crops. Therefore, this study proposes field programmable gate array (FPGA)-based hardware solutions and real-time kinematic GPS (RTK-GPS) to develop a real-time crop-monitoring system that can address the limitations of current PA technologies. Our proposed system uses high-accuracy RTK and real-time FPGA-based image-processing (RFIP) devices for data collection, geotagging real-time field data via Python and a camera. The acquired images are processed to extract metadata then visualized as a heat map on Google Maps, indicating green area intensity based on romaine lettuce leafage. The RFIP system showed a strong correlation (R2 = 0.9566) with a reference system and performed well in field tests, providing a Lin’s concordance correlation coefficient (CCC) of 0.8292. This study demonstrates the potential of the developed system to address current PA limitations by providing real-time, accurate data for immediate decision making. In the future, this proposed system will be integrated with autonomous farm equipment to further enhance sustainable farming practices, including real-time crop health monitoring, yield assessment, and crop disease detection.
Selecting uniform and healthy seedlings is important to ensure that a certain level of production can be reliably achieved in a plant factory. The objectives of this study were to investigate the potential of non-destructive image analysis for predicting the leaf area and shoot fresh weight of lettuce and to determine the feasibility of using a simple image analysis to select robust seedlings that can produce a uniform and dependable yield of lettuce in a plant factory. To vary the range of the leaf area and shoot fresh weight of lettuce seedlings, we applied two- and three-day irrigation intervals during the period of seedling production and calculated the projected canopy size (PCS) from the top-view images of the lettuce seedlings, although there were no significant growth differences between the irrigation regimes. A high correlation was identified between the PCS and shoot fresh weight for the lettuce seedlings during the period of seedling production, with a coefficient of determination exceeding 0.8. Therefore, the lettuce seedlings were classified into four grades (A–D) based on their PCS values calculated at transplanting. In the early stages of cultivation after transplanting, there were differences in the lettuce growth among the four grades; however, at the harvest (28 days after transplanting), there was no significant difference in the lettuce yield between grades A–C, with the exception of grade D. The lettuce seedlings in grades A–C exhibited the anticipated yield (150 g/plant) at the harvest time. In the correlation between the PCS and leaf area or the shoot fresh weight of lettuce during the cultivation period after transplanting and the entire cultivation period, the R2 values were higher than 0.9, confirming that PCS can be used to predict lettuce growth with greater accuracy. In conclusion, we demonstrated that the PCS calculation from the top-view images, a straightforward image analysis technique, can be employed to non-destructively and accurately predict lettuce leaf area and shoot fresh weight, and the seedlings with the potential to yield above a certain level after transplanting can be objectively and accurately selected based on PCS.
Introduction Soil-borne pathogens cause considerable crop losses and food insecurity in smallholder systems of sub-Saharan Africa. Soil and crop testing is critical for estimating pathogen inoculum levels and potential for disease development, understanding pathogen interactions with soil nutrient and water limitations, as well as for developing informed soil health and disease management decisions. However, formal laboratory analyses and diagnostic services for pathogens are often out of reach for smallholder farmers due to the high cost of testing and a lack of local laboratories. Methods To address this challenge, we assessed the performance of a suite of simplified soil bioassays to screen for plant parasitic nematodes (e.g., Meloidogyne , Pratylenchus ) and other key soil-borne pathogens ( Pythium and Fusarium ). We sampled soils from on-farm trials in western Kenya examining the impact of distinct nutrient inputs (organic vs. synthetic) on bean production. Key soil health parameters and common soil-borne pathogens were evaluated using both simple bioassays and formal laboratory methods across eleven farms, each with three nutrient input treatments (66 samples in total). Results and discussion The soil bioassays, which involved counting galls on lettuce roots and lesions on soybean were well correlated with the abundance of gall forming ( Meloidogyne ) and root lesion nematodes (e.g., Pratylenchus ) recovered in standard laboratory-based extractions. Effectiveness of a Fusarium bioassay, involving the counting of lesions on buried bean stems, was verified via sequencing and a pathogenicity test of cultured Fusarium strains. Finally, a Pythium soil bioassay using selective media clearly distinguished pathogen infestation of soils and infected seeds. When examining management impact on nematode communities, soils amended with manure had fewer plant parasites and considerably more bacterivore and fungivore nematodes compared to soils amended with synthetic N and P. Similarly, Pythium presence was 35% lower in soils amended with manure, while the Fusarium assays indicated 23% higher Fusarium infection in plots with amended manure. Our findings suggest that relatively simple bioassays can be used to help farmers assess soil-borne pathogens in a timely manner, with minimal costs, thus enabling them to make informed decisions on soil health and pathogen management.
Kulaporn Boonyaves · Mervin Chun-Yi Ang · Minkyung Park · Jianqiao Cui · Duc Thinh Khong · Gajendra Pratap Singh · Volodymyr B. Koman · Xun Gong · Thomas K. Porter · Nam-Hai Chua · Daisuke Urano · Michael S. Strano
Gibberellins (GAs) are a class of phytohormones, important for plant growth, and very difficult to distinguish because of their similarity in chemical structures. Herein, we develop the first nanosensors for GAs by designing and engineering polymer-wrapped single-walled carbon nanotubes (SWNTs) with unique corona phases that selectively bind to bioactive GAs, GA 3 and GA 4 , triggering near-infrared (NIR) fluorescence intensity changes. Using a new coupled Raman/NIR fluorimeter that enables self-referencing of nanosensor NIR fluorescence with its Raman G-band, we demonstrated detection of cellular GA in Arabidopsis, lettuce, and basil roots. The nanosensors reported increased endogenous GA levels in transgenic Arabidopsis mutants that overexpress GA and in emerging lateral roots. Our approach allows rapid spatiotemporal detection of GA across species. The reversible sensor captured the decreasing GA levels in salt-treated lettuce roots, which correlated remarkably with fresh weight changes. This work demonstrates the potential for nanosensors to solve longstanding problems in plant biotechnology.
In recent years, precision agriculture, driven by scientific monitoring, precise management, and efficient use of agricultural resources, has become the direction for future agricultural development. The precise identification and assessment of phenotypes, which serve as external representations of a crop's growth, development, and genetic characteristics, are crucial for the realization of precision agriculture. Applications surrounding phenotypic indices also provide significant technical support for optimizing crop cultivation management and advancing smart agriculture, contributing to the efficient and high-quality development of precision agriculture.This paper focuses on lettuce and employs common nutritional stress conditions during growth as experimental settings. By collecting RGB images throughout the lettuce's complete growth cycle, we developed a deep learning-based computational model to tackle key issues in the lettuce's growth and precisely identify and assess phenotypic indices. We discovered that some phenotypic indices, including custom ones defined in this study, are representative of the lettuce's growth status. By dynamically monitoring the changes in phenotypic traits during growth, we quantitatively analyzed the accumulation and evolution of phenotypic indices across different growth stages. On this basis, a predictive model for lettuce growth and development was trained.The model incorporates MSE, SSIM, and perceptual loss, significantly enhancing the predictive accuracy of the lettuce growth images and phenotypic indices. The model trained with the reconstructed loss function outperforms the original model, with the SSIM and PSNR improving by 1.33% and 10.32%, respectively. The model also demonstrates high accuracy in predicting lettuce phenotypic indices, with an average error less than 0.55% for geometric indices and less than 1.7% for color and texture indices. Ultimately, it achieves intelligent monitoring and management throughout the lettuce's life cycle, providing technical support for high-quality and efficient lettuce production.
The survival and growth of young plants hinge on various factors, such as seed quality and environmental conditions. Assessing seedling potential/vigor for a robust crop yield is crucial but often resource-intensive. This study explores cost-effective imaging techniques for rapid evaluation of seedling vigor, offering a practical solution to a common problem in agricultural research. In the first phase, nine lettuce ( Lactuca sativa ) cultivars were sown in trays and monitored using chlorophyll fluorescence imaging thrice weekly for two weeks. The second phase involved integrating embedded computers equipped with cameras for phenotyping. These systems captured and analyzed images four times daily, covering the entire growth cycle from seeding to harvest for four specific cultivars. All resulting data were promptly uploaded to the cloud, allowing for remote access and providing real-time information on plant performance. Results consistently showed the 'Muir' cultivar to have a larger canopy size and better germination, though 'Sparx' and 'Crispino' surpassed it in final dry weight. A non-linear model accurately predicted lettuce plant weight using seedling canopy size in the first study. The second study improved prediction accuracy with a sigmoidal growth curve from multiple harvests ( R 2 = 0.88, RMSE = 0.27, p < 0.001). Utilizing embedded computers in controlled environments offers efficient plant monitoring, provided there is a uniform canopy structure and minimal plant overlap.
The decline in seed quality over time due to natural aging or mishandling requires assessing seed vigor for resilience in adverse conditions. Accelerated aging (AA) methods simulate seed deterioration by subjecting seeds to high temperatures and humidity. Saturated salt accelerated aging (SSAA) is an AA method adopted for small seeds like lettuce ( Lactuca sativa ). In this study, we subjected seeds of two lettuce cultivars ('Muir' and 'Bauer') to SSAA by sealing them in a box containing 40 g/100 mL of a sodium chloride (NaCl) solution in a dark growth chamber at 41 °C for 24, 48, and 72 h with a control. We monitored their vigor using embedded computer cameras, tracking the projected canopy size (PCS) daily from sowing to harvest. The cultivar 'Muir' exhibited consistent PCS values across the treatments, while 'Bauer' showed PCS variations, with notable declines after prolonged aging. The germination rates dropped significantly after 48 and 72 h of SSAA. A nonlinear regression model revealed a strong relationship between PCS and shoot dry weight across harvests and cultivars ( R 2 = 0.93, RMSE = 0.15, p < 0.001). The research found that the projected canopy size and shoot dry weight increased over time with significant differences in treatments for the cultivar 'Bauer' but not for 'Muir,' with the canopy size being a strong predictor of dry weight and no significant impact from the SSAA treatments. This study highlights cultivar-specific responses to aging and demonstrates the efficacy of our imaging tool in predicting lettuce dry weight despite treatment variations. Understanding how aging affects different lettuce varieties is crucial for seed management and crop sustainability.
Salinity stress is a common challenge in plant growth, impacting seed quality, germination, and general plant health. Sodium chloride (NaCl) ions disrupt membranes, causing ion leakage and reducing seed viability. Gibberellic acid (GA 3 ) treatments have been found to promote germination and mitigate salinity stress on germination and plant growth. 'Bauer' and 'Muir' lettuce ( Lactuca sativa ) seeds were soaked in distilled water (control), 100 mM NaCl, 100 mM NaCl + 50 mg/L GA 3 , and 100 mM NaCl + 150 mg/L GA 3 in Petri dishes and kept in a dark growth chamber at 25 °C for 24 h. After germination, seedlings were monitored using embedded cameras, capturing red, green, and blue (RGB) images from seeding to final harvest. Despite consistent germination rates, 'Bauer' seeds treated with NaCl showed reduced germination. Surprisingly, the 'Muir' cultivar's final dry weight differed across treatments, with the NaCl and high GA 3 concentration combination yielding the poorest results ( p 3 applications in improving germination rates. However, at elevated concentrations, it induced excessive hypocotyl elongation and pale seedlings, posing challenges for two-dimensional imaging. Nonetheless, a sigmoidal regression model using projected canopy size accurately predicted dry weight across growth stages and cultivars, emphasizing its reliability despite treatment variations ( R 2 = 0.96, RMSE = 0.11, p < 0.001).
Lettuce is a globally important cash crop, valued by consumers for its nutritional content and pleasant taste. However, there is limited research on the changes in the growth indicators of lettuce during its growth period in domestic settings. Quality assessment primarily relies on subjective evaluations, resulting in significant variability. This study focused on hydroponically grown lettuce during the rosette stage and investigated the patterns of changes in the indicators and spectral curves over time. By employing spectral preprocessing and selecting characteristic wavelengths, three models were developed to predict the indicators. The results showed that the optimal model structures were S_G-UVE-PLSR (SSC and vitamin C) and Nor-CARS-PLSR (moisture content). The PLSR models achieved prediction set correlation coefficients of 0.8648, 0.8578, and 0.8047, with residual prediction deviations of 1.9685, 1.9568, and 1.6689, respectively. The optimal models were integrated into a portable device, using real-time analysis software written in Matlab2021a, for the prediction of the physicochemical indicators of lettuce during the rosette stage. The results demonstrated prediction set correlation coefficients of 0.8215, 0.8472, and 0.7671, with root mean square errors of prediction of 0.5348, 1.5813, and 2.3347 for a sample size of 180. The small discrepancies between the predicted and actual values indicate that the developed device can meet the requirements for real-time detection.
Introduction Indoor agriculture, especially plant factories, becomes essential because of the advantages of cultivating crops yearly to address global food shortages. Plant factories have been growing in scale as commercialized. Developing an on-site system that estimates the fresh weight of crops non-destructively for decision-making on harvest time is necessary to maximize yield and profits. However, a multi-layer growing environment with on-site workers is too confined and crowded to develop a high-performance system. This research developed a machine vision-based fresh weight estimation system to monitor crops from the transplant stage to harvest with less physical labor in an on-site industrial plant factory. Methods A linear motion guide with a camera rail moving in both the x-axis and y-axis directions was produced and mounted on a cultivating rack with a height under 35 cm to get consistent images of crops from the top view. Raspberry Pi4 controlled its operation to capture images automatically every hour. The fresh weight was manually measured eleven times for four months to use as the ground-truth weight of the models. The attained images were preprocessed and used to develop weight prediction models based on manual and automatic feature extraction. Results and discussion The performance of models was compared, and the best performance among them was the automatic feature extraction-based model using convolutional neural networks (CNN; ResNet18). The CNN-based model on automatic feature extraction from images performed much better than any other manual feature extraction-based models with 0.95 of the coefficients of determination (R 2 ) and 8.06 g of root mean square error (RMSE). However, another multiplayer perceptron model (MLP_2) was more appropriate to be adopted on-site since it showed around nine times faster inference time than CNN with a little less R 2 (0.93). Through this study, field workers in a confined indoor farming environment can measure the fresh weight of crops non-destructively and easily. In addition, it would help to decide when to harvest on the spot.
Abstract Lettuce are vegetables with a high commercial value and a short cycle production, which requires precise managements to guarantee a profitable production. For this, the nutritional providing is an important factor, highlining the nitrogen, a macronutrient considered essential for the proper development of plants due to its participation in the composition of the main plant molecules, such as chlorophyll. In this sense, non-destructive strategies to monitor the balance nutrition is fundamental to avoid deficiency or excess of nutrients during the lettuce cycling, and the agriculture 4.0, brings to light new technological means to achieve this goal. This work aimed to verify the potential to use a portable spectrophotometer to estimate N shoot in lettuce, as a non-destructive, precise, quickly, cheap, waste free option to be operated by the lettuce farmers, on farm, for the N monitoring, based on the destructive laboratory analysis for N shoot and chlorophyll, and the widely used sensor chlorophyll meter (SPAD). For this, a greenhouse experiment was conducted with lettuce cropped under N levels input, corresponding to 0, 25, 50, 75, 100 and 125% of the lettuce recommendation. The estimation for N in the lettuce shoot was performed by destructive analysis: i) shoot N; ii) chlorophyll and non-destructive analysis: iii) chlorophyll meter SPAD and iv) a portable spectrophotometer. Non-destructive measurements were performed 3 times at: 20, 23 and 26 days after transplanting (DAT). The lettuce harvest occurred 28 DAT, to determine N shoot uptake and chlorophyll, using a destructive way, and lettuce production. The estimation of N shoot in lettuce using the portable spectrophotometer showed a high correlation to the standard destructive laboratory analysis and the chlorophyll meter (SPAD), showing high explanation of the data and so a high potential to estimate N shoot in lettuce using the proposed portable spectrophotometer as an optimum non-destructive, precise, quickly, cheap, waste free option to be operated by the lettuce farmers, on farm, for the N monitoring.
Lettuce (Lactuca sativa) is a leafy vegetable that provides a valuable source of phytonutrients for a healthy human diet. The assessment of plant growth and composition is vital for determining crop yield and overall quality; however, classical laboratory analyses are slow and costly. Therefore, new, less expensive, more rapid, and non-destructive approaches are being developed, including those based on (hyper)spectral reflectance. Additionally, it is important to determine how plant phenotypes respond to fertilizer treatments and whether these differences in response can be detected from analyses of hyperspectral image data. In the current study, we demonstrate the suitability of hyperspectral imaging in combination with machine learning models to estimate the content of chlorophyll (SPAD), anthocyanins (ACI), glucose, fructose, sucrose, vitamin C, β-carotene, nitrogen (N), phosphorus (P), potassium (K), dry matter content, and plant fresh weight. Five classification and regression machine learning models were implemented, showing high accuracy in classifying the lettuces based on the applied fertilizers treatments and estimating nutrient concentrations. To reduce the input (predictor data, i.e., hyperspectral data) dimension, 13 principal components were identified and applied in the models. The implemented artificial neural network models of the machine learning algorithm demonstrated high accuracy (r = 0.85 to 0.99) in estimating fresh leaf weight, and the contents of chlorophyll, anthocyanins, N, P, K, and β-carotene. The four applied classification models of machine learning demonstrated 100% accuracy in classifying the studied baby leaf lettuces by phenotype when specific fertilizer treatments were applied.
Abstract Seed germination research has evolved over the years, increasingly incorporating technology. Recent advances in phenotyping platforms have increased the accessibility of high throughput phenotyping technologies to more labs, leading to valuable insights into germination biology. These platforms benefit researchers by limiting manual labor and increasing the temporal resolution of imaging. Each of the platforms developed presents unique benefits and challenges, from scalability to price to computing resources. Performing experiments involving thousands of seeds remains a daunting task due to the limitations of current phenotyping platforms and image analysis pipelines. To overcome these challenges, we introduce SPENCER (Seed Phenotype Evaluation and Germination Curve Estimation Robot), a high-throughput phenotyping platform. SPENCER accommodates 32 rectangular petri plates, capable of assessing up to 8000 Arabidopsis seeds per experiment. Our design allows for high quality images while maintaining optimal humidity, crucial for precise germination assessment over longer experiments. The image analysis workflow incorporates advanced image analysis using semantic segmentation models trained for Arabidopsis and lettuce, providing researchers with accessible, reproducible, and efficient tools. We applied SPENCER to investigate the relative roles of DELAY OF GERMINATION 1 (DOG1) and abscisic acid (ABA) in Arabidopsis dormancy. DOG1 mutants exhibited rapid germination, whereas ANT application had a greater impact on the slower-germinating Ler ecotype. Our findings suggest that DOG1 plays a significant role in dormancy, particularly in non-dormant accessions, while ABA’s influence is more pronounced under stress conditions. Additionally, we explored germination uniformity, another agriculurally relevant trait, observing parallels with germination timing. SPENCER offers a powerful and accessible tool for dissecting complex biological traits in conjunction with chemical and genetic manipulations. Its scalability and versatility make it suitable for large-scale genetic and chemical germination screens.
Lettuce (Lactuca sativa) is a leafy vegetable that provides a valuable source of phytonutrients for a healthy human diet. Assessment of plant growth and composition is vital for determining crop yield and overall quality, however, classical laboratory analyses are slow and costly. Therefore, new, less expensive, more rapid, and non-destructive approaches are being developed, including those based on (hyper)spectral reflectance. Additionally, it is also important to determine how plant phenotypes respond to fertilizer treatments and whether these differences in response can be detected from analyses of hyperspectral image data. In the current study, we demonstrate the suitability of hyperspectral imaging in combination with machine learning models to estimate the content of chlorophyll (SPAD), anthocyanins (ACI), glucose, fructose, sucrose, vitamin C, β-carotene, N, P, K, dry matter content, and plant fresh weight. The implemented five classification and regression machine learning models showed high accuracy in classifying the lettuces by the applied fertilizers treatments and estimating nutrient concentrations. To reduce the input (predictor data, i.e., hyperspectral data) dimension, 13 principal components were found and applied in models. The implemented artificial neural network models of the machine learning algorithm demonstrated high accuracy (r = 0.85 ... 0.99) in estimating fresh leaf weight, and contents of chlorophyll, anthocyanins, N, P, K, and β-carotene. The four applied classification models of machine learning demonstrated 100% accuracy in classifying the studied baby leaf lettuces by phenotype when certain fertilizer treatments were applied.
Mohamed Mydin M. Abdul Kader · Muhammad Naufal Mansor · Zol Bahri Razali · Wan Azani Mustafa · Ahmad Anas Nagoor Gunny · Samsul Setumin · Muhammad Khusairi Osman · Mohaiyedin Idris · Muhammad Firdaus Akbar · Wan Muhamad Faris Naim Muhami Farid · Muhammad Zubir Zainol · Nor Syamina Sharifful Mizam
Malaysia is currently driven to become another most developed country in the world. Among other priority sector is Food Sustainability. Along the process, our vegetable supply-demand keeps increasing by year. Compared to traditional systems, closed systems or its other name called hydroponic is getting more important for plant production, with artificial light which has many potential advantages, including better quality transplants, shorter production time and less resource use. To gain full profit from it, the quality of vegetables needs to be controlled efficiently. Climate conditions, especially temperature and light intensity, have a significant impact on vegetable growth and yield, as well as nutritional quality. Plant growth and development are influenced by a variety of environmental factors, the most important one is light intensity. Among the problems to be tackled in this research are plant growth manual observation, light intensity variation and abundance of growth-related data to be evaluated manually. Therefore, to solve these problems, the specific type of vegetable used here is lettuce. The proposed methods are, observation of plant growth conducted automatically round the clock in intervals of 15 minutes for the whole month (estimated mature period of lettuce), using images captured. At the same time, the proposed light intensity which is red & white to the ratio of 2:1 (optimum ratio recommended by previous researchers) will be used. The issue of data to be evaluated manually will be solved using Artificial Neural Network (ANN) architecture, in specific Deep Learning. Concisely, the results & analysis shows the research is successfully developed for plant growth monitoring by using artificial neural network which, reached 80% to 90% accuracy in the training and validation session that made the architecture sufficient for determining the growth of the said vegetable. This is indeed foreseen, will highly assist the farmer in better monitoring the growth rate of the plant.
The agriculture sector is the most important contributor to expanding economies and people because of the vital role it plays in providing high-quality food. It is possible for plant diseases to cause significant decreases in food production as well as the extinction of endangered species. Improving food production quality and minimizing economic losses can be achieved by early identification of plant diseases utilizing reliable or automated detection techniques. Deep learning has recently made great strides in improving the accuracy of object detection and picture categorization systems. The authors of this research proposed a new method for detecting plant diseases; they called it the Learning Network for Disease Identification (LNDI), and they cross-validated it with the Convolutional Neural Network (CNN). Differentiating between diseases in different plant species is a key goal of this research. Several plant kinds, including as fruits and vegetables, wheat, raisins, sugarcane, and lettuce have been incorporated into the system's development process. A wide variety of herbal ailments can also be diagnosed by the computer. Using a large dataset consisting of photos of diseased and healthy plant leaves, the specialists trained deep learning models to detect and distinguish between various plant illnesses and those that went unnoticed. Biological research and agricultural institutes are only two of the many potential uses for plant leaf disease detection. Research into plant leaf disease detection is necessary because it has the potential to improve crop monitoring by automatically identifying disease signs on plant leaves as they emerge.
In recent years, precision agriculture, driven by scientific monitoring, precise management, and efficient use of agricultural resources, has become the direction for future agricultural development. The precise identification and assessment of phenotypes, which serve as external representations of a crop's growth, development, and genetic characteristics, are crucial for the realization of precision agriculture. Applications surrounding phenotypic indices also provide significant technical support for optimizing crop cultivation management and advancing smart agriculture, contributing to the efficient and high-quality development of precision agriculture.This paper focuses on lettuce and employs common nutritional stress conditions during growth as experimental settings. By collecting RGB images throughout the lettuce's complete growth cycle, we developed a deep learning-based computational model to tackle key issues in the lettuce's growth and precisely identify and assess phenotypic indices. We discovered that some phenotypic indices, including custom ones defined in this study, are representative of the lettuce's growth status. By dynamically monitoring the changes in phenotypic traits during growth, we quantitatively analyzed the accumulation and evolution of phenotypic indices across different growth stages. On this basis, a predictive model for lettuce growth and development was trained.The model incorporates MSE, SSIM, and perceptual loss, significantly enhancing the predictive accuracy of the lettuce growth images and phenotypic indices. The model trained with the reconstructed loss function outperforms the original model, with the SSIM and PSNR improving by 1.33% and 10.32%, respectively. The model also demonstrates high accuracy in predicting lettuce phenotypic indices, with an average error less than 0.55% for geometric indices and less than 1.7% for color and texture indices. Ultimately, it achieves intelligent monitoring and management throughout the lettuce's life cycle, providing technical support for high-quality and efficient lettuce production.
Crop growth monitoring is pivotal in optimizing management strategies and maximizing greenhouse production. Traditionally, crop monitoring is carried out manually, which makes it unfeasible to collect data daily to get actionable insights for high yield. This study presents an innovative, non-destructive approach to predict lettuce growth parameters, including leaf area, fresh weight, dry weight, plant diameter, and plant height. The proposed methodology capitalizes on the capabilities of a semantic segmentation model, specifically, a lightweight DeepLabv3 + network that integrates MobileNetv2. This model showcases exceptional performance with a mean IoU score of 0.9979, accuracy of 0.9985, and a segmentation speed of 0.075fps. Furthermore, the study assesses the performance of the deep learning regression model in predicting lettuce phenotypic parameters, achieving R² values of 0.968, 0.953, 0.943, 0.906, and 0.965 for fresh weight, leaf area, dry weight, plant diameter, and plant height, respectively. To underscore the model's robustness, it is subjected to validation under various treatment conditions, encompassing variations in nutrients and temperature. Our findings revealed that the treatment involving high nutrient temperature and medium N contents (Temp: 30 °C, Nitrogen: 150 ppm) yielded the highest fresh and dry weights. These validations substantiate the efficacy of the predictive model for hydroponic lettuce, and this innovative approach holds promise for data aggregation and predictive analytics to assist growers in decision-making for resource optimization.
Background The phenotypic traits of leaves are the direct reflection of the agronomic traits in the growth process of leafy vegetables, which plays a vital role in the selection of high-quality leafy vegetable varieties. The current image-based phenotypic traits extraction research mainly focuses on the morphological and structural traits of plants or leaves, and there are few studies on the phenotypes of physiological traits of leaves. The current research has developed a deep learning model aimed at predicting the total chlorophyll of greenhouse lettuce directly from the full spectrum of hyperspectral images. Results A CNN-based one-dimensional deep learning model with spectral attention module was utilized for the estimate of the total chlorophyll of greenhouse lettuce from the full spectrum of hyperspectral images. Experimental results demonstrate that the deep neural network with spectral attention module outperformed the existing standard approaches, including partial least squares regression (PLSR) and random forest (RF), with an average R 2 of 0.746 and an average RMSE of 2.018. Conclusions This study unveils the capability of leveraging deep attention networks and hyperspectral imaging for estimating lettuce chlorophyll levels. This approach offers a convenient, non-destructive, and effective estimation method for the automatic monitoring and production management of leafy vegetables.
Diets consisting of greater quantity/diversity of phytochemicals are correlated with reduced risk of disease. This understanding guides policy development increasing awareness of the importance of consuming fruits, grains, and vegetables. Enacted policies presume uniform concentrations of phytochemicals across crop varieties regardless of production/harvesting methods. A growing body of research suggests that concentrations of phytochemicals can fluctuate within crop varieties. Improved awareness of how cropping practices influence phytochemical concentrations are required, guiding policy development improving human health. Reliable, inexpensive laboratory equipment represents one of several barriers limiting further study of the complex interactions influencing crop phytochemical accumulation. Addressing this limitation our study validated the capacity of a low-cost Reflectometer ($500) to measure phytochemical content in selected crops, against a commercial grade laboratory spectrophotometer. Our correlation results ranged from r 2 = 0.81 for protein in wheat and oats to r 2 = 0.99 for polyphenol content in lettuce in both the Reflectometer and laboratory spectrophotometer assessment, suggesting the Reflectometer provides an accurate accounting of phytochemical content within evaluated crops. Repeatability evaluation demonstrated good reproducibility of the Reflectometer to assess crop phytochemical content. Additionally, we confirmed large variation in phytochemical content within specific crop varieties, suggesting that cultivar is but one of multiple drivers of phytochemical accumulation. Our findings indicate dramatic nutrient variations could exist across the food supply, a point whose implications are not well understood. Future studies should investigate the interactions between crop phytochemical accumulation and farm management practices that influence specific soil characteristics.
Reproduction assets foundThe paper explicitly states that all Bionutrient Institute data (reflectometer/spectrometer phytochemical measurements used in this study) are publicly available in the authors' GitLab repository, and the authors' data-processing pipeline code is also publicly hosted on GitLab.Dataset · publicAll data derived from the Bionutrient Institute methods are available publicly from our repository: https://gitlab.com/our-sci/bionutrient-institute/dataset . The data used in this manuscript covers samples submitted up to 7/31/2022.Open asset ↗our-sci/bionutrient-institute/datasetlines:156-212Code · publicAn automated data pipeline was built using SurveyStacks API’s to merge data from each completed survey and mongoDB scripts ( https://gitlab.com/our-sci/real-food-campaign/lab-data-review-dashboard/-/tree/main ) calculated measurement outcomes.Open asset ↗our-sci/real-food-campaign/lab-data-review-dashboardlines:132-143Code / dataset availability confirmedEurope PMC · checked 14 Sept 2026
Introduction The cold stress is one of the most important factors for affecting production throughout year, so effectively evaluating frost damage is great significant to the determination of the frost tolerance in lettuce. Methods We proposed a high-throughput method to estimate lettuce FDI based on remote sensing. Red-Green-Blue (RGB) and multispectral images of open-field lettuce suffered from frost damage were captured by Unmanned Aerial Vehicle platform. Pearson correlation analysis was employed to select FDI-sensitive features from RGB and multispectral images. Then the models were established for different FDI-sensitive features based on sensor types and different groups according to lettuce colors using multiple linear regression, support vector machine and neural network algorithms, respectively. Results and discussion Digital number of blue and red channels, spectral reflectance at blue, red and near-infrared bands as well as six vegetation indexes (VIs) were found to be significantly related to the FDI of all lettuce groups. The high sensitivity of four modified VIs to frost damage of all lettuce groups was confirmed. The average accuracy of models were improved by 3% to 14% through a combination of multisource features. Color of lettuce had a certain impact on the monitoring of frost damage by FDI prediction models, because the accuracy of models based on green lettuce group were generally higher. The MULTISURCE-GREEN-NN model with R 2 of 0.715 and RMSE of 0.014 had the best performance, providing a high-throughput and efficient technical tool for frost damage investigation which will assist the identification of cold-resistant green lettuce germplasm and related breeding.
Reproduction assets found保存済みの本文根拠を更新済みルールで再検証し、公開資産1件を確認しました。Code · publicData, models, or codes generated or used in the course of the study are available on GitHub at https://github.com/kwcnmm/predict-FDI .Open asset ↗kwcnmm/predict-FDIlines:908-915Plant phenotyping relevance matchEurope PMC · checked 15 Sept 2026
Early detection of pathogenic fungi in controlled environment areas can prevent major food production losses. Grey mould caused by Botrytis cinerea is often detected as an infection on lettuce. This paper explores the use of vegetation indices for early detection and monitoring of grey mould on lettuce under different lighting conditions in controlled environment chambers. The aim was focused on the potential of using vegetation indices for the early detection of grey mould and on evaluating their changes during disease development in lettuce grown under different lighting conditions. The experiment took place in controlled environment chambers, where day/night temperatures were 21 ± 2/17 ± 2 °C, a 16 h photoperiod was established, and relative humidity was 70 ± 10% under different lighting conditions: high-pressure sodium (HPS) and light-emitting diode (LED) lamps. Lettuces were inoculated by 7-day-old fungus Botrytis cinerea isolate at the BBCH 21. As a control, non-inoculated lettuces were grown under HPS and LEDs (non-inoculated). Then, the following were evaluated: Anthocyanin Reflectance Index 2 (ARI2); Carotenoid Reflectance Index 2 (CRI2); Structure Intensive Pigment Index (SIPI); Flavanol Reflectance Index (FRI); Greenness (G); Greenness 2 (G2); Redness (R); Blue (B); Blue Green Index 2 (BGI2); Browning Index 2 (BRI2); Lichtenthaler Index 1 (LIC1); Pigment Specific Simple Ratio (PSSRa and PSSRb); Gitelson and Merzlyak (GM1 and GM2); Zarco Tejada-Miller Index (ZMI); Normalized Difference Vegetation Index (NDVI); Simple Ratio (SR); Red-Eye Vegetation Stress Index (RVSI); Photochemical Reflectance Index (PRI); Photochemical Reflectance Index 515 (PRI515); Water Band Index (WBI); specific disease index for individual study (fD); Healthy Index (HI); Plant Senescence Reflectance (PSRI); Vogelmann Red Edge Index (VREI1); Red Edge Normalized Difference Vegetation Index (RENDVI); and Modified Red Edge Simple Ratio (MRESRI). Our results showed that the PSRI and fD vegetation indices significantly detected grey mould on lettuce grown under both lighting systems (HPS and LEDs) the day after inoculation. The results conclusively affirmed that NDVI, PSRI, HI, fD, WBI, RVSI, PRI, PRI515, CRI2, SIPI, chlorophyll index PSSRb, and coloration index B were identified as the best indicators for Botrytis cinerea infection on green-leaf lettuce ( Lactuca sativa L. cv Little Gem) at the early stage of inoculated lettuce's antioxidative response against grey mould with a significant increase in chlorophyll indices.
This article presents an integrated system that uses the capabilities of unmanned aerial vehicles (UAVs) to perform a comprehensive crop analysis, combining qualitative and quantitative evaluations for efficient agricultural management. A convolutional neural network-based model, Detectron2, serves as the foundation for detecting and segmenting objects of interest in acquired aerial images. This model was trained on a dataset prepared using the COCO format, which features a variety of annotated objects. The system architecture comprises a frontend and a backend component. The frontend facilitates user interaction and annotation of objects on multispectral images. The backend involves image loading, project management, polygon handling, and multispectral image processing. For qualitative analysis, users can delineate regions of interest using polygons, which are then subjected to analysis using the Normalized Difference Vegetation Index (NDVI) or Optimized Soil Adjusted Vegetation Index (OSAVI). For quantitative analysis, the system deploys a pre-trained model capable of object detection, allowing for the counting and localization of specific objects, with a focus on young lettuce crops. The prediction quality of the model has been calculated using the AP (Average Precision) metric. The trained neural network exhibited robust performance in detecting objects, even within small images.
Computer vision provides a real-time, non-destructive, and indirect way of horticultural crop yield estimation. Deep learning helps improve horticultural crop yield estimation accuracy. However, the accuracy of current estimation models based on RGB (red, green, blue) images does not meet the standard of a soft sensor. Through enriching more data and improving the RGB estimation model structure of convolutional neural networks (CNNs), this paper increased the coefficient of determination (R2) by 0.0284 and decreased the normalized root mean squared error (NRMSE) by 0.0575. After introducing a novel loss function mean squared percentage error (MSPE) that emphasizes the mean absolute percentage error (MAPE), the MAPE decreased by 7.58%. This paper develops a lettuce fresh weight estimation method through the multi-modal fusion of RGB and depth (RGB-D) images. With the multimodal fusion based on calibrated RGB and depth images, R2 increased by 0.0221, NRMSE decreased by 0.0427, and MAPE decreased by 3.99%. With the novel loss function, MAPE further decreased by 1.27%. A MAPE of 8.47% helps to develop a soft sensor for lettuce fresh weight estimation.
Demand for sustainable and safe raw agricultural commodities is growing rapidly worldwide. Reducing the risk of foodborne illnesses associated with fresh produce is a task which the industry and academic researchers have been struggling with for many years. There is an immediate need to devise a non‐invasive optical detection system to monitor the food‐borne pathogens on the leaf surface. The detection of foodborne pathogens on leafy produce is performed often too late because of the invasive techniques used to evaluate the pathogen colonization. Use of deep ultraviolet fluorescence (DUVF) sensing and visible–near infrared multispectral imaging (MSI) has previously been used to monitor plant interactions against both biotic and abiotic stress regimes. Using the patho‐system that we developed to monitor Salmonella sp. and Listeria sp. ingression in leafy greens such as lettuce/spinach, we show that plant response in terms of fluctuation of chlorophyll pigments post‐Salmonella/Listeria treatment is rapid. We also show that the mode of application of Salmonella/Listeria via foliar or root supplementation changes the ChlA response. Our data also reveals that the plant sentinel response in terms of early photosynthetic response may be critical to detect food‐borne pathogens on leafy greens. MSI demonstrated that plant stress was detectable and proportional to the bacterial inoculation rate on plants. Our research may lead to implementation of better strategies and technology to increase yield and reduce risks associated with contamination of foodborne bacterial pathogens.
Plant phenotyping has been widely studied as an effective and powerful tool for analyzing crop status and growth. However, the traditional phenotyping (i.e., manual) is time-consuming and laborious, and the various types of growing structures and limited room for systems hinder phenotyping on a large and high-throughput scale. In this study, a low-cost high-throughput phenotyping system that can be flexibly applied to diverse structures of growing beds with reliable spatial–temporal continuities was developed. The phenotyping system was composed of a low-cost phenotype sensor network with an integrated Raspberry Pi board and camera module. With the distributed camera sensors, the system can provide crop imagery information over the entire growing bed in real time. Furthermore, the modularized image-processing architecture supports the investigation of several phenotypic indices. The feasibility of the system was evaluated for Batavia lettuce grown under different light periods in a container-type plant factory. For the growing lettuces under different light periods, crop characteristics such as fresh weight, leaf length, leaf width, and leaf number were manually measured and compared with the phenotypic indices from the system. From the results, the system showed varying phenotypic features of lettuce for the entire growing period. In addition, the varied growth curves according to the different positions and light conditions confirmed that the developed system has potential to achieve many plant phenotypic scenarios at low cost and with spatial versatility. As such, it serves as a valuable development tool for researchers and cultivators interested in phenotyping.
One of the main challenges facing the development of aquaponics is disease control, due on one hand to the fact that plants cannot be treated with chemicals because they can lead to mortality in cultured fish. The aim of this study was to apply the visible-near-infrared spectroscopy and vegetation index approach to test aquaponically cultivated lettuce ( Lactuca sativa L.) infected with different fungal pathogens ( Aspergillus niger , Fusarium oxysporum , and Alternaria alternata ). The lettuces on the third leaf formation were placed in tanks (with dimensions 1 m/0.50 m/0.35 m) filled up with water from the aquaponics system every second day. In this study, we included reference fungal strains Aspergillus niger NBIMCC 3252, Fusarium oxysporum NBIMCC 125, and Alternaria alternata NBIMCC 109. Diffuse reflectance spectra of the leaves of lettuce were measured directly on the plants using a USB4000 spectrometer in the 450-1100 nm wavelength range. In near-infrared spectral range, the reflectance values of infected leaves are lower than those of the control, which indicates that some changes in cell structures occurred as a result of the fungal infection. All three investigated pathogens had a statistically significant effect on leaf water content and water band index. Vegetative indices such as Chlorophyll Absorption in Reflectance Index (CARI), Modified chlorophyll absorption in reflectance index (MCARI), Plant Senescence Reflectance Index (PSRI), Red Edge Index (REI2), Red Edge Index (REI3), and Water band index (WBI) were found to be effective in distinguishing infected plants from healthy ones, with WBI demonstrating the greatest reliability.
Our laboratory at MIT has been interested over the past few years in new techniques to facilitate the transfer of chemical information from living organisms, specifically plants, animals and humans, for applications ranging from precision agriculture to precision medicine. This presentation will discuss recent advances on this topic. As tool towards this end, fluorescent nanosensors hold the potential to revolutionize life sciences and medicine. However, their adaptation and translation into the in vivo environment is fundamentally hampered by unfavourable tissue scattering and intrinsic autofluorescence. Here we develop wavelength-induced frequency filtering (WIFF) whereby the fluorescence excitation wavelength is modulated across the absorption peak of a nanosensor, allowing the emission signal to be separated from the autofluorescence background, increasing the desired signal relative to noise, and internally referencing it to protect against artefacts. Using highly scattering phantom tissues, an SKH1-E mouse model and other complex tissue types, we show that WIFF improves the nanosensor signal-to-noise ratio across the visible and near-infrared spectra up to 52-fold. This improvement enables the ability to track fluorescent carbon nanotube sensor responses to riboflavin, ascorbic acid, hydrogen peroxide and a chemotherapeutic drug metabolite for depths up to 5.5 ± 0.1 cm when excited at 730 nm and emitting between 1,100 and 1,300 nm, even allowing the monitoring of riboflavin diffusion in thick tissue. As an application, nanosensors aided by WIFF detect the chemotherapeutic activity of temozolomide transcranially at 2.4 ± 0.1 cm through the porcine brain without the use of fibre optic or cranial window insertion. The ability of nanosensors to monitor previously inaccessible in vivo environments will be important for life-sciences research, therapeutics and medical diagnostics. Also towards this overall objective, our laboratory at MIT has been interested in exploring the relatively new interface between living plants and non-biological nanostructures to impart the former with new and enhanced functions, which we call Plant Nanobionics. We have developed a theory of subcellular uptake and kinetic trapping of a wide range of nanoparticles, validated in-vivo in living plants. Confocal visible and near infrared fluorescent microscopy and single particle tracking of Gold-Cystein-AF405 (GNP-Cys-AF405), Streptavidin-Quantum Dot (SA-QD), Dextran and Poly(acrylic acid) nanoceria, and various polymer-wrapped SWCNT, including lipid-PEG-SWCNT, chitosan-SWCNT and (AT)15-SWCNT, were used to demonstrate that particle size and the magnitude, but not the sign, of the zeta potential are key in determining whether a particle is spontaneously and kinetically trapped within chloroplasts or the cytosol. We develop a mathematical model of this Lipid Exchange Envelope Penetration (LEEP) mechanism, which agrees well with observations of this size and zeta potential dependence. As an application, we rationally designed a chitosan-complexed single-walled carbon nanotube (SWNT) as nanocarriers to selectively deliver plasmid DNA (pDNA) to chloroplasts of different plant species without external biolistic or chemical aid. We demonstrate chloroplast-targeted transgene delivery and expression in living mature arugula (Eruca sativa) and watercress (Nasturitium officinale) plants in planta and in isolated Arabidopsis thaliana mesophyll protoplasts. Another application of nanoparticles and nanotechnology to plant sciences is in the form of biochemical sensors that operate in planta and across diverse species. Using non-destructive optical nanosensors, we find that the spatial and temporal H2O2 concentration immediately post-wounding follows a simple logistic waveform for six dicot plant species: lettuce (Lactuca sativa), arugula (Eruca sativa), spinach (Spinacia oleracea), strawberry blite (Blitum capitatum), sorrel (Rumex acetosa), and Arabidopsis thaliana, ranked in order of wave speed from 0.44 to 3.10 cm/min. The H2O2 wave tracks the concomitant surface potential wave measured electrochemically for the series of plants. We show that the plant NADPH oxidase RbohD, glutamate receptor-like channels (GLR3.3 and GLR3.6) are all critical to the propagation of the H2O2 waveform upon wounding. Our findings highlight the utility of a new type of nanosensor probe that is species-independent and capable of real-time, spatial and temporal biochemical measurements in planta.
Lettuce is one of the most widely planted leafy vegetables in plant factories. The lack of trace elements in nutrient solutions has caused huge losses to the lettuce industry. Non-obvious symptoms of trace element deficiency, the inconsistent size of the characteristic areas, and the difficulty of extraction in different growth stages are three key problems affecting lettuce deficiency symptom identification. In this study, a batch of cream lettuce (lactuca sativa) was planted in the plant factory, and its nutrient elements were artificially controlled. We collected images of the lettuce at different growth stages, including all nutrient elements and three nutrient-deficient groups (potassium deficiency, calcium deficiency, and magnesium deficiency), and performed feature extraction analysis on images of different defects. We used traditional algorithms (k-nearest neighbor, support vector machine, random forest) and lightweight deep-learning models (ShuffleNet, SqueezeNet, andMobileNetV2) for classification, and we compared different feature extraction methods (texture features, color features, scale-invariant feature transform features). The experiment shows that, under the optimal feature extraction method (color), the random-forest recognition results are the best, with an accuracy rate of 97.6%, a precision rate of 97.9%, a recall rate of 97.4%, and an F1 score of 97.6%. The accuracies of all three deep-learning models exceed 99.5%, among which ShuffleNet is the best, with the accuracy, precision, recall, and F1 score above 99.8%. It also uses fewer floating-point operations per second and less time. The proposed method can quickly identify the trace elements lacking in lettuce, and it can provide technical support for the visual recognition of the disease patrol robot in the plant factory.
Three-dimensional measurement is a high-throughput method that can record a large amount of information. Three-dimensional modelling of plants has the possibility to not only automate dimensional measurement, but to also enable visual assessment to be quantified, eliminating ambiguity in human judgment. In this study, we have developed new methods that could be used for the morphological analysis of plants from the information contained in 3D data. Specifically, we investigated characteristics that can be measured by scale (dimension) and/or visual assessment by humans. The latter is particularly novel in this paper. The characteristics that can be measured on a scale-related dimension were tested based on the bounding box, convex hull, column solid, and voxel. Furthermore, for characteristics that can be evaluated by visual assessment, we propose a new method using normal vectors and local curvature (LC) data. For these examinations, we used our highly accurate all-around 3D plant modelling system. The coefficient of determination between manual measurements and the scale-related methods were all above 0.9. Furthermore, the differences in LC calculated from the normal vector data allowed us to visualise and quantify the concavity and convexity of leaves. This technique revealed that there were differences in the time point at which leaf blistering began to develop among the varieties. The precise 3D model made it possible to perform quantitative measurements of lettuce size and morphological characteristics. In addition, the newly proposed LC-based analysis method made it possible to quantify the characteristics that rely on visual assessment. This research paper was able to demonstrate the following possibilities as outcomes: (1) the automation of conventional manual measurements, and (2) the elimination of variability caused by human subjectivity, thereby rendering evaluations by skilled experts unnecessary.
Compact and automated sensing systems are needed to monitor plant health for NASA's controlled-environment space crop production. A new hyperspectral system was designed for early detection of plant stresses using both reflectance and fluorescence imaging in visible and near-infrared (VNIR) wavelength range (400-1000 nm). The prototype system mainly includes two LED line lights providing VNIR broadband and UV-A (365 nm) light for reflectance and fluorescence measurement, respectively, a line-scan hyperspectral camera, and a linear motorized stage with a travel range of 80 cm. In an overhead sensor-to-sample arrangement, the stage translates the lights and camera over the plants to acquire reflectance and fluorescence images in sequence during one cycle of line-scan imaging. System software was developed using LabVIEW to realize hardware parameterization, data transfer, and automated imaging functions. The imaging unit was installed in a plant growth chamber at NASA Kennedy Space Center for health monitoring studies for pick-and-eat salad crops. A preliminary experiment was conducted to detect plant drought stress for twelve Dragoon lettuce samples, of which half were well-watered and half were under-watered while growing. A machine learning method using an optimized discriminant classifier based on VNIR reflectance spectra generated classification accuracies over 90% for the first four days of the stress treatment, showing great potential for early detection of the drought stress on lettuce leaves before any visible symptoms and size differences were evident. The system is promising to provide useful information for optimization of growth environment and early mitigation of stresses in space crop production.
Efficiently analyzing the relationship between plant phenotypes, quality, and resistance remains challenging. In this study, deep learning models based on hyperspectral data and time-series phenotypes from the high-throughput plant phenotyping (HTPP) platform were proposed to predict quality attributes of lettuce under water stress, including SSC, pH value, nitrate (NO₃–), and calcium (Ca²⁺). First, deep learning models were developed using the Inception module and raw hyperspectral data to non-destructively predict the above quality attributes. In addition, partial least squares regression (PLSR) and support vector regression (SVR) were used to develop prediction models to evaluate performance of the Inception module. Second, the residual and attention modules were implemented to enhance performance of the Inception module. Third, time-series phenotypes were fed into four recurrent neural networks (RNNs), such as TimeDistributed (TD), long short-term memory (LSTM), Gated Recurrent Unit (GRU), and Bidirectional RNN (BRNN) and combined with the optimal deep learning models based on hyperspectral data to enhance prediction precision. The optimal performance of the Inception-residual-attention-TD model was achieved with Rₚ² of 0.8900 and 0.9435 for SSC and NO₃–, respectively. The Inception-residual-TD model with Rₚ² of 0.9583 provided the most accurate pH value prediction. With Rₚ² of 0.8716, the Inception-attention-LSTM model provided the most accurate prediction of Ca²⁺. Meanwhile, the Inception-residual-TD model was used to detect water stress, producing an Accuracyₚ of 98.86%. The Inception-residual model based on pixel-wise hyperspectral data was used to visualize the spatial distribution of pH value, and the distribution map was used to detect early water stress. The results indicate that deep learning models can use hyperspectral data and time-series phenotypes to predict lettuce quality attributes and water stress in a non-destructive manner.
3D measurement is a high-throughput method that can record a large amount of information. In this study, we have developed new methods that could be used for morphological analysis of plants from the information contained in 3D data. Specifically, we investigated characteristics that can be measured by scale (dimension) and/or visual assessment by humans. The characteristics that can be measured on a scale-related dimension were tested based on the bounding box, convex hull, column solid, and voxel. Furthermore, for characteristics that can be evaluated by visual assessment, we propose a new method using normal vectors and local curvature (LC) data. For these examinations, we used our highly accurate all-around 3D plant modelling system. The correlation coefficients between manual measurements and the scale-related methods were all above 0.9. In particular, the differences in LC calculated from the normal vector data allowed us to visualize and quantify the concavity and convexity of leaves. Furthermore, we also found a difference in the time point at which leaf blistering began to develop among the cultivars. The precise 3D model made it possible to perform quantitative measurements of lettuce size and morphological characteristics. In addition, the newly proposed LC-based analysis method made it possible to quantify the characteristics that rely on visual assessment.
Advanced precision agriculture requires the objective measurement of the structural and functional properties of plants. Biochemical profiles in leaves can differ depending on plant growing conditions. By quantitatively detecting these changes, farm production processes can be optimized to achieve high-yield, high-quality, and nutrient dense agricultural products. To enable the rapid and non-destructive detection on site, this study demonstrates the development of a new custom-designed portable handheld Vis-NIR spectrometer that collects leaf reflectance spectra, wirelessly transfers the spectral data through Bluetooth, and provides both raw spectral data and processed information. The spectrometer has two preprogramed methods: anthocyanin and chlorophyll quantification. Anthocyanin content of red and green lettuce estimated with the new spectrometer showed an excellent correlation coefficient of 0.84 with those determined by a destructive gold standard biochemical method. The differences in chlorophyll content were measured using leaf senescence as a case study. Chlorophyll Index calculated with the handheld spectrometer gradually decreased with leaf age as chlorophyll degrades during the process of senescence. The estimated chlorophyll values were highly correlated with those obtained from a commercial fluorescence-based chlorophyll meter with a correlation coefficient of 0.77. The developed portable handheld Vis-NIR spectrometer could be a simple, cost-effective, and easy to operate tool that can be used to non-invasively monitor plant pigment and nutrient content efficiently.
Effective management of plant essential nutrients is necessary for hydroponically grown lettuce to achieve high yields and maintain production. This study investigated in situ hyperspectral imaging of hydroponic lettuce for predicting nutrient concentrations and identifying nutrient deficiencies for: nitrogen (N), phosphorous (P), potassium (K), calcium (Ca), magnesium (Mg), and sulphur (S). A greenhouse study was conducted using ‘Salanova Green’ lettuce grown with controlled solution treatments with varying macronutrient fertility rates of 0, 8, 16, 32, 64, and 100% each for N, P, K, Ca, Mg, and S. Plants were imaged using a hyperspectral line scanner at six and eight weeks after transplanting; then, plant tissues were sampled, and nutrient concentrations measured. Partial least squares regression (PLSR) models were developed to predict nutrient concentrations for each nutrient individually (PLS1) and for all six nutrient concentrations (PLS2). Several binary classification models were also developed to predict nutrient deficiencies. The PLS1 and PLS2 models predicted nutrient concentrations with Coefficient of Determination (R²) values from 0.60 to 0.88 for N, P, K, and S, while results for Ca and Mg yielded R² values of 0.12–0.34, for both harvest dates. Similarly, plants deficient in N, P, K, and S were classified more accurately compared to plants deficient in Ca and Mg for both harvest dates, with F1 values (F-scores) ranging from 0.71 to 1.00, with the exception of K which had F1 scores of 0.40–0.67. Overall, results indicate that both leaf tissue nutrient concentration and nutrient deficiencies can be predicted using hyperspectral data collected for whole plants.
Currently, the concept of plant capture efficiency is not quantitatively considered in the evaluation of off-target drift for the purposes of pesticide risk assessment in the United States. For on-target pesticide applications, canopy capture efficiency is managed by optimizing formulations or tank-mixing with adjuvants to maximize retention of spray droplets. These efforts take into consideration the fact that plant species have diverse morphology and surface characteristics, and as such will retain varying levels of applied pesticides. This work aims to combine plant surface wettability potential, spray droplet characteristics, and plant morphology into describing the plant capture efficiency of drifted spray droplets. In this study, we used wind tunnel experiments and individual plants grown to 10-20 cm to show that at two downwind distances and with two distinct nozzles capture efficiency for sunflower (Helianthus annuus L.), lettuce (Lactuca sativa L.), and tomato (Solanum lycopersicum L.) is consistently higher than rice (Oryza sativa L.), peas (Pisum sativum L). and onions (Allium cepa L.), with carrots (Daucus carota L.) showing high variability and falling between the two groups. We also present a novel method for three-dimensional modeling of plants from photogrammetric scanning and use the results in the first known computational fluid dynamics simulations of drift capture efficiency on plants. The mean simulated drift capture efficiency rates were within the same order of magnitude of the mean observed rates of sunflower and lettuce, and differed by one to two orders for rice and onion. We identify simulating the effects of surface roughness on droplet behavior, and the effects of wind flow on plant movement as potential model improvements requiring further species-specific data collection.
Plant stress responses involve a suite of genetically encoded mechanisms triggered by real-time interactions with their surrounding environment. Although sophisticated regulatory networks maintain proper homeostasis to prevent damage, the tolerance thresholds to these stresses vary significantly among organisms. Current plant phenotyping techniques and observables must be better suited to characterize the real-time metabolic response to stresses. This impedes practical agronomic intervention to avoid irreversible damage and limits our ability to breed improved plant organisms. Here, we introduce a sensitive, wearable electrochemical glucose-selective sensing platform that addresses these problems. Glucose is a primary plant metabolite, a source of energy produced during photosynthesis, and a critical molecular modulator of various cellular processes ranging from germination to senescence. The wearable-like technology integrates a reverse iontophoresis glucose extraction capability with an enzymatic glucose biosensor that offers a sensitivity of 22.7 nA/(μM·cm 2 ), a limit of detection (LOD) of 9.4 μM, and a limit of quantification (LOQ) of 28.5 μM. The system's performance was validated by subjecting three different plant models (sweet pepper, gerbera, and romaine lettuce) to low-light and low-high temperature stresses and demonstrating critical differential physiological responses associated with their glucose metabolism. This technology enables non-invasive, non-destructive, real-time, in-situ, and in-vivo identification of early stress response in plants and provides a unique tool for timely agronomic management of crops and improving breeding strategies based on the dynamics of genome-metabolome-phenome relationships.
Anthocyanins provide blue, red, and purple color to fruits, vegetables, and flowers. Due to their benefits for human health and aesthetic appeal, anthocyanin content in crops affects consumer preference. Rapid, low-cost, and non-destructive phenotyping of anthocyanins is not well developed. Here, we introduce the normalized difference anthocyanin index (NDAI), which is based on the optical properties of anthocyanins: high absorptance in the green and low absorptance in the red part of the spectrum. NDAI is determined as (I red - I green )/(I red + I green ), where I is the pixel intensity, a measure of reflectance. To test NDAI, leaf discs of two red lettuce ( Lactuca sativa ) cultivars ‘Rouxai’ and ‘Teodore’ with wide range of anthocyanin concentrations were imaged using a multispectral imaging system and the red and green images were used to calculate NDAI. NDAI and other commonly used indices for anthocyanin quantification were evaluated by comparing to with the measured anthocyanin concentration (n = 50). Statistical results showed that NDAI has advantages over other indices in terms of prediction of anthocyanin concentrations. Canopy NDAI, obtained using multispectral canopy imaging, was correlated (n = 108, R 2 = 0.73) with the anthocyanin concentrations of the top canopy layer, which is visible in the images. Comparison of canopy NDAI from multispectral images and RGB images acquired using a Linux-based microcomputer with color camera, showed similar results in the prediction of anthocyanin concentration. Thus, a low-cost microcomputer with a camera can be used to build an automated phenotyping system for anthocyanin content.
In this study, we investigated the use of artificial intelligence algorithms (AIAs) in combination with VIS-NIR-SWIR hyperspectroscopy for the classification of eleven lettuce plant varieties. For this purpose, a spectroradiometer was utilized to collect hyperspectral data in the VIS-NIR-SWIR range, and 17 AIAs were applied to classify lettuce plants. The results showed that the highest accuracy and precision were achieved using the full hyperspectral curves or the specific spectral ranges of 400-700 nm, 700-1300 nm, and 1300-2400 nm. Four models, AdB, CN2, G-Boo, and NN, demonstrated exceptional R 2 and ROC values, exceeding 0.99, when compared between all models and confirming the hypothesis and highlighting the potential of AIAs and hyperspectral fingerprints for efficient, precise classification and pigment phenotyping in agriculture. The findings of this study have important implications for the development of efficient methods for phenotyping and classification in agriculture and the potential of AIAs in combination with hyperspectral technology. To advance our understanding of the capabilities of hyperspectroscopy and AIs in precision agriculture and contribute to the development of more effective and sustainable agriculture practices, further research is needed to explore the full potential of these technologies in different crop species and environments.
Reproduction assets found保存済みの本文根拠を更新済みルールで再検証し、公開資産1件を確認しました。Supplement · publicThe following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/plants12061333/s1 . Table S1. Descriptive analysis parameters of lettuce varieties. Pigment of leaves expressed by leaf area (mg m −2 ), mass (mg g −1 ), and volume (mL L −1 ) ( n = 132); Table S2. STEPW and VIPs by wavelengths selected according to classified algorithm-based ANOVA and information gain ratio ( p < 0.001) by band range spectroscopy from reflectance leaves.Open asset ↗lines:80-115Code / dataset availability confirmedEurope PMC · Crossref · checked 7 Sept 2026
Recent studies indicate that food demand will increase by 35-56% over the period 2010-2050 due to population increase, economic development, and urbanization. Greenhouse systems allow for the sustainable intensification of food production with demonstrated high crop production per cultivation area. Breakthroughs in resource-efficient fresh food production merging horticultural and AI expertise take place with the international competition "Autonomous Greenhouse Challenge". This paper describes and analyzes the results of the third edition of this competition. The competition's goal is the realization of the highest net profit in fully autonomous lettuce production. Two cultivation cycles were conducted in six high-tech greenhouse compartments with operational greenhouse decision-making realized at a distance and individually by algorithms of international participating teams. Algorithms were developed based on time series sensor data of the greenhouse climate and crop images. High crop yield and quality, short growing cycles, and low use of resources such as energy for heating, electricity for artificial light, and CO 2 were decisive in realizing the competition's goal. The results highlight the importance of plant spacing and the moment of harvest decisions in promoting high crop growth rates while optimizing greenhouse occupation and resource use. In this paper, images taken with depth cameras (RealSense) for each greenhouse were used by computer vision algorithms (Deepabv3+ implemented in detectron2 v0.6) in deciding optimum plant spacing and the moment of harvest. The resulting plant height and coverage could be accurately estimated with an R 2 of 0.976, and a mIoU of 98.2, respectively. These two traits were used to develop a light loss and harvest indicator to support remote decision-making. The light loss indicator could be used as a decision tool for timely spacing. Several traits were combined for the harvest indicator, ultimately resulting in a fresh weight estimation with a mean absolute error of 22 g. The proposed non-invasively estimated indicators presented in this article are promising traits to be used towards full autonomation of a dynamic commercial lettuce growing environment. Computer vision algorithms act as a catalyst in remote and non-invasive sensing of crop parameters, decisive for automated, objective, standardized, and data-driven decision making. However, spectral indexes describing lettuces growth and larger datasets than the currently accessible are crucial to address existing shortcomings between academic and industrial production systems that have been encountered in this work.
Reproduction assets foundThe paper's complete challenge dataset (climate time-series and annotated lettuce crop images used for the computer vision phenotyping) is published open access on 4TU.ResearchData, cited both in the Data Availability Statement and in reference 56.Dataset · public3rd Autonomous Greenhouse Challenge-Real Challenge Data Climate and Images Dataset: 4TU.ResearchData 2023 Available online: https://data.4tu.nl/articles/dataset/3rd_Autonomous_Greenhouse_Challenge_Online_Challenge_Lettuce_Images/15023088Open asset ↗4TU.ResearchData · 15023088lines:853-968Code / dataset availability confirmedEurope PMC · Crossref · checked 7 Sept 2026
Gonzalez EM, Zarei A, Hendler N, Simmons T, Zarei A, Demieville J, Strand R, Rozzi B, Calleja S, Ellingson H, Cosi M, Davey S, Lavelle DO, Truco MJ, Swetnam TL, Merchant N, Michelmore RW, Lyons E, Pauli D.
As phenomics data volume and dimensionality increase due to advancements in sensor technology, there is an urgent need to develop and implement scalable data processing pipelines. Current phenomics data processing pipelines lack modularity, extensibility, and processing distribution across sensor modalities and phenotyping platforms. To address these challenges, we developed PhytoOracle (PO), a suite of modular, scalable pipelines for processing large volumes of field phenomics RGB, thermal, PSII chlorophyll fluorescence 2D images, and 3D point clouds. PhytoOracle aims to ( i ) improve data processing efficiency; ( ii ) provide an extensible, reproducible computing framework; and ( iii ) enable data fusion of multi-modal phenomics data. PhytoOracle integrates open-source distributed computing frameworks for parallel processing on high-performance computing, cloud, and local computing environments. Each pipeline component is available as a standalone container, providing transferability, extensibility, and reproducibility. The PO pipeline extracts and associates individual plant traits across sensor modalities and collection time points, representing a unique multi-system approach to addressing the genotype-phenotype gap. To date, PO supports lettuce and sorghum phenotypic trait extraction, with a goal of widening the range of supported species in the future. At the maximum number of cores tested in this study (1,024 cores), PO processing times were: 235 minutes for 9,270 RGB images (140.7 GB), 235 minutes for 9,270 thermal images (5.4 GB), and 13 minutes for 39,678 PSII images (86.2 GB). These processing times represent end-to-end processing, from raw data to fully processed numerical phenotypic trait data. Repeatability values of 0.39-0.95 (bounding area), 0.81-0.95 (axis-aligned bounding volume), 0.79-0.94 (oriented bounding volume), 0.83-0.95 (plant height), and 0.81-0.95 (number of points) were observed in Field Scanalyzer data. We also show the ability of PO to process drone data with a repeatability of 0.55-0.95 (bounding area).
Reproduction assets foundThe paper's Code and Data Availability statements provide explicit public URLs for the authors' PhytoOracle processing code, ML training-data preparation scripts, trained model training code, and the season-10 lettuce benchmarking dataset (raw RGB/thermal/PSII images and point clouds) hosted on CyVerse.Dataset · publicThe datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: https://datacommons.cyverse.org/browse/iplant/home/shared/phytooracle/season_10_lettuce_yr_2020Open asset ↗iplant/home/shared/phytooracle/season_10_lettuce_yr_2020lines:640-662Code · publicThe automation script and data processing repositories can be accessed at: http://github.com/phytooracleOpen asset ↗github.com/phytooraclelines:640-662Code · publicThe Python scripts used to prepare RGB training data can be accessed here: http://github.com/phytooracle/automation/blob/main/ml/collect_rgb_data.pyOpen asset ↗github.com/phytooracle/automationlines:640-662Code · publicThe Python script used to prepare thermal training data can be accessed here: http://github.com/phytooracle/automation/blob/main/ml/collect_flir_data.pyOpen asset ↗github.com/phytooracle/automationlines:640-662Code · publicThe Python script used to prepare 3D-derived images can be found here: http://github.com/phytooracle/3d_heat_map/blob/main/3d_heat_map.pyOpen asset ↗github.com/phytooracle/3d_heat_maplines:640-662Plant phenotyping relevance matchEurope PMC · checked 15 Sept 2026
Published1 Mar 2023Computers and Electronics in Agriculture.
Lettuce growth and light energy consumption in a plant factory with artificial lighting (PFAL) were studied, and whole plant photosynthetic rate (ACO₂) and light use efficiency (LUE) data were obtained on different days after planting (DAP) under different photosynthetic photon flux densities (PPFDs). Genetic algorithm‐support vector regression (GA-SVR) was used to construct the ACO₂ and LUE prediction models. The coefficient of determination (R²) between the predicted and measured values of the ACO₂ model was 0.97 and the root mean square error (RMSE) was 0.42 μmol·mol⁻¹·plant⁻¹·min⁻¹, and R² between the predicted and measured values of the LUE model was 0.97 and RMSE was 0.36%. The ACO₂ and LUE prediction models were used as the objective functions, and multi-objective search was performed by the non-dominated sorting genetic algorithm II (NSGA-II) and the distance-based knee point detection method were used to obtained the optimal equilibrium solution for different DAPs. The optimal equilibrium solutions were used as the basis to establish the light regulation model based on lettuce DAP with R² of 0.99. To validate the effect of model regulation, a lettuce light regulation system was built using an artificial climate chamber for a 30-day system validation. The results showed that compared with the traditional quantitative light supplementation method, the dry matter of model regulation significantly increased by 39.23% and 29.48% compared with quantitative PPFD150 (μmol·m⁻²·s⁻¹) and PPFD200 (μmol·m⁻²·s⁻¹). Model regulation increased the number of total light quanta consumed by 1.39% over PPFD150 and decreased by 23.96% over PPFD200; however, plant productivity increased by 35.35% and 33.14%, respectively. Model regulation significantly reduced the number of light quanta consumed per unit mass of lettuce production by 24.35% for PPFD150 and 41.54% for PPFD200, and LUE of light-emitting diode energy into dry matter was significantly increased by 33.49% and 75.09%. Therefore, the light regulation model based on multi-objective optimization in this study could improve crop yield and increase LUE.
Abstract Estimation of crop water stress index (CWSI) and leaf area index (LAI) over large-irrigation schemes requires the use of cutting-edge technologies. Combinations of remote sensing techniques with ground-truth data have become available for use at the catchment level. These approaches allow us to estimate actual evapotranspiration and have the capability of monitoring crop water status and saving irrigation water in water-scarce regions. This study was conducted in the eastern Mediterranean Region of Turkiye. Fully distributed CWSI maps were generated and we assessed the relationship between CWSI and LAI for some specific crops in the winter season of 2021. Landsat 7 and 8 data were used and meteorological data were acquired from two stations in the study area. ‘Mapping Evapotranspiration at high Resolution with Internalized Calibration’ methodology was applied to estimate the energy balance components. CWSI maps displayed spatiotemporal changes in tandem with crop-type variations. Consequently, results presented a high correlation (r = 0.95 and r = 0.99 for wheat and lettuce, respectively) between CWSI and LAI, a moderate correlation (r = 0.44) for potatoes in the winter season. Thus, by utilizing remotely sensed data, the CWSI values would be directly estimated without requiring any in situ measurements of the canopy and air temperature over-irrigation scheme.
Unmanned ground vehicles (UGV) have attracted much attention in crop phenotype monitoring due to their lightweight and flexibility. This paper describes a new UGV equipped with an electric slide rail and point cloud high-throughput acquisition and phenotype extraction system. The designed UGV is equipped with an autopilot system, a small electric slide rail, and Light Detection and Ranging (LiDAR) to achieve high-throughput, high-precision automatic crop point cloud acquisition and map building. The phenotype analysis system realized single plant segmentation and pipeline extraction of plant height and maximum crown width of the crop point cloud using the Random sampling consistency (RANSAC), Euclidean clustering, and k-means clustering algorithm. This phenotyping system was used to collect point cloud data and extract plant height and maximum crown width for 54 greenhouse-potted lettuce plants. The results showed that the correlation coefficient (R2) between the collected data and manual measurements were 0.97996 and 0.90975, respectively, while the root mean square error (RMSE) was 1.51 cm and 4.99 cm, respectively. At less than a tenth of the cost of the PlantEye F500, UGV achieves phenotypic data acquisition with less error and detects morphological trait differences between lettuce types. Thus, it could be suitable for actual 3D phenotypic measurements of greenhouse crops.
Advanced precision agriculture requires the objective measurement of the structural and functional properties of plants. Biochemical profiles in leaves can differ depending on plant growing conditions. By quantitatively detecting these changes, farm production processes can be optimized to achieve high-yield, high-quality, and nutrient dense agricultural products. To enable the rapid and non-destructive detection on site, this study demonstrates the development of a new custom-designed portable handheld Vis-NIR spectrometer that collects leaf reflectance spectra, wirelessly transfers the spectral data through Bluetooth, and provides both raw spectral data and processed information. The spectrometer has two preprogramed methods: anthocyanin and chlorophyll quantification. Anthocyanin content of red and green lettuce estimated with the new spectrometer showed an excellent correlation coefficient of 84% with those determined by a destructive gold standard biochemical method. The differences in chlorophyll content were measured using leaf senescence as a case study. Chlorophyll Index calculated with the handheld spectrometer gradually decreased with leaf age as chlorophyll degrades during the process of senescence. The estimated chlorophyll values were highly correlated with those obtained from a commercial fluorescence-based chlorophyll meter with a correlation coefficient of 77%. The developed portable handheld Vis-NIR spectrometer could be a simple, cost-effective, and easy to operate tool that can be used to non-invasively monitor plant pigment and nutrient content efficiently.
Gibberellins (GAs) are a class of phytohormones, important for plant growth, and very difficult to distinguish because of their similarity in chemical structures. Herein, we develop the first nanosensors for GAs by designing and engineering polymer-wrapped single-walled carbon nanotubes (SWNTs) with unique corona phases that selectively bind to bioactive GAs, GA 3 and GA 4 , triggering near-infrared (NIR) fluorescence intensity changes. Using a new coupled Raman/NIR fluorimeter that enables self-referencing of nanosensor NIR fluorescence with its Raman G-band, we demonstrated detection of cellular GA in Arabidopsis , lettuce, and basil roots. The nanosensors reported increased endogenous GA levels in transgenic Arabidopsis mutants that overexpress GA and in emerging lateral roots. Our approach allows rapid spatiotemporal detection of GA across species. The reversible sensor captured the decreasing GA levels in salt-treated lettuce roots, which correlated remarkably with fresh weight changes. This work demonstrates the potential for nanosensors to solve longstanding problems in plant biotechnology.
To measure gas exchange at the whole plant (WP) level, design solutions were provided and prototypes of gas-exchange systems (GESs) were tested to carry out shoot and rootzone analyses in plants of different sizes. A WP-GES for small herbaceous plants was tested on the ability to maximize the net assimilation rate of CO₂ in lettuce plants grown either under blue-red light or upon full spectrum artificial light. A WP-GES for large woody plants was tested during an experiment describing the drought stress inhibition of grapevine transpiration and photosynthesis. Technical advances pointed to optimize: i) the choice of cuvette material and its technical configuration to allow hermetic isolation of the interface shoot-rootzone, to avoid contamination between the two compartments, and to allow climate control of both shoot and rootzone cuvettes, ii) accurate measurements of the mass air-flow entering both cuvettes, and iii) an adequate homogenization of the cuvette air volume for stable and accurate detection of CO₂ and H₂O concentration in cuvettes before and after CO₂ and H₂O contamination of the air volumes exerted by plant organs.
Jongyoon Kim · Yu Kyeong Shin · Yunsu Nam · Jun Gu Lee · Ji-Hoon Lee
LettuceLeafPhysiological trait estimationGrowth / development / phenologyStress response / tolerance
Abstract Polarimetry is a powerful characterization technique that uses a wealth of information from electromagnetic waves, including polarization. Using the rich information provided by polarimetry, it is being actively studied in biomedical fields such as cancer and tumor diagnosis. Despite its importance and potential in agriculture, polarimetry for living plants has not been well studied. A Stokes polarimetric imaging system was built to determine the correlation between the polarization states of the light passing through the leaf and the growth states of lettuce. The Stokes parameter s 3 associated with circular polarization increased over time and was strongly correlated with the growth of lettuce seedlings. In the statistical analysis, the distribution of s 3 followed the generalized extreme value (GEV) probability density function. Salt stress retarded plant growth, and the concentration of treated sodium chloride (NaCl) showed a negative correlation with the location parameter μ of GEV. The clear correlation reported here will open the possibility of polarization measurements on living plants, enabling real-time monitoring of plant health.
Assessment of plant traits (phenotyping) is central to modern advanced techniques of plant sciences and accelerated breeding of crop plants, including fruit crops, for improving productivity and stress resilience. Hyperspectral reflectance imaging is an emerging method allowing to capture a vast amount of the structural, biochemical, and phenological information about plants. The advent of low-cost hyperspectrometers made this method affordable for a broad community of plant scientists. However, extraction of sensible information from reflectance images is hindered by the complexity of plant optical properties, especially when they are measured in the field. We propose using reflectance indices (Plant Senescence Reflectance Index, PSRI; Anthocyanin Reflectance Index, ARI; and spectral deconvolution) previously developed for remote sensing of vegetation and point-based reflectometers to infer the spatially resolved information on plant development and biochemical composition using lettuce (Lactuca sativa L.) leaves and ripening apple (Malus × domestica Borkh.) fruit as the model. Specifically, the proposed approach enables capturing data on distribution of chlorophylls and primary carotenoids as well as secondary carotenoids (both linked with fruit ripening and leaf senescence during plant development) as well as the information on spatial distribution of anthocyanins (known as stress pigments) over the plant surface. We argue that the proposed approach would enrich the phenotype assessments made on the base of reflectance image analysis with valuable information on plant physiological condition, stress acclimation state, and the progression of the plant development.
Rising urban food demand is being addressed by plant factories, which aim at producing quality food in closed environment with optimised use of resources. The efficiency of these new plant production systems could be further increased by automated control of plant health and nutritious composition during cultivation, allowing for increased produce value and closer match between plant needs and treatment application with potential energy savings. We hypothesise that certain leaf pigments, including chlorophylls, carotenoids and anthocyanins, which are responsive to light, may be good indicator of plant performance and related healthy compounds composition and, that the combination of leaf spectroscopy and mathematical modelling will allow monitoring of plant cultivation through noninvasive estimation of leaf pigments. Plants of two lettuce cultivars (a green- and a red-leaf) were cultivated in hydroponic conditions for 18 days under white light spectrum in climate controlled growth chamber. After that period, plant responses to white light spectrum ('W') with differing blue wavelengths ('B', 420 - 450 nm) percentage (15% 'B15', and 40% 'B40') were investigated for a 14 days period. The two light spectral treatments were applied at photon flux densities (PFDs) of 160 and 240 µmol m -2 s -1 , resulting in a total of four light treatments (160WB15, 160WB40, 240WB15, 240WB40). Chlorophyll a fluorescence measurements and assessment of foliar pigments, through destructive ( in vitro ) and non-destructive ( in vivo ) spectrophotometry, were performed at 1, 7 and 14 days after treatment initiation. Increase in measured and estimated pigments in response to WB40 and decrease in chlorophyll:carotenoid ratio in response to higher PFD were found in both cultivars. Cultivar specific behavior in terms of specific pigment content stimulation in response to time was observed. Content ranges of modelled and measured pigments were comparable, though the correlation between both needs to be improved. In conclusion, leaf pigment estimation may represent a potential noninvasive and real-time technique to monitor, and control, plant growth and nutritious quality in controlled environment agriculture.
Hydroponic lettuce has been widely cultivated in plant factory and desiring for mechanical harvesting and packing. Sorting of hydroponic lettuce must be carried out before packing. Information perception and image processing of hydroponic lettuce is a crucial technology to develop a robotic sorting system. In this study, DeepLabV3+ models of deep learning technologies were employed with four backbones of ResNet-50, ResNet-101, Xception-65, and Xception-71 to design a vision system of segmenting abnormal leaves (yellow, withered, and decay leaves) of hydroponic lettuce. Two weights assignation methods, i.e., median frequency weights (MFW) and uniform weights (UW), were incorporated into DeepLabV3+ and compared for performance. Results showed that models trained by UW were better than that of MFW assignation method. ResNet-101 had the best segmentation performance in UW assignation method with pixel accuracy of 99.24% and mIoU of 0.8326. In terms of speed, ResNet-50 had the fast segmentation speeds with 154.0 ms per image. This study provided object detection methodology for automatic sorting device of hydroponic lettuce.
To measure gas exchange at the whole plant (WP) level, design solutions were provided and prototypes of gas-exchange systems (GESs) were tested to carry out shoot and rootzone analyses in plants of different sizes. A WP-GES for small herbaceous plants was tested on the ability to maximize the net assimilation rate of CO 2 in lettuce plants grown either under blue-red light or upon full spectrum artificial light. A WP-GES for large woody plants was tested during an experiment describing the drought stress inhibition of grapevine transpiration and photosynthesis. Technical advances pointed to optimize: i) the choice of cuvette material and its technical configuration to allow hermetic isolation of the interface shoot-rootzone, to avoid contamination between the two compartments, and to allow climate control of both shoot and rootzone cuvettes, ii) accurate measurements of the mass air-flow entering both cuvettes, and iii) an adequate homogenization of the cuvette air volume for stable and accurate detection of CO 2 and H 2 O concentration in cuvettes before and after CO 2 and H 2 O contamination of the air volumes exerted by plant organs.
Wild species of lettuce ( Lactuca sp.) are thought to have first been domesticated for oilseed contents to provide seed oil for human consumption. Although seed morphology is an important trait contributing to oilseed in lettuce, the underlying genetic mechanisms remain elusive. Since lettuce seeds are small, a manual phenotypic determination required for a genetic dissection of such traits is challenging. In this study, we built and applied an instance segmentation-based seed morphology quantification pipeline to measure traits in seeds generated from a cross between the domesticated oilseed type cultivar ‘Oilseed’ and the wild species ‘UenoyamaMaruba’ in an automated manner. Quantitative trait locus (QTL) mapping following ddRAD-seq revealed 11 QTLs linked to 7 seed traits (area, width, length, length-to-width ratio, eccentricity, perimeter length, and circularity). Remarkably, the three QTLs with the highest LOD scores, qLWR-3.1 , qECC-3.1 , and qCIR-3.1 , for length-to-width ratio, eccentricity, and circularity, respectively, mapped to linkage group 3 (LG3) around 161.5 to 214.6 Mb, a region previously reported to be associated with domestication traits from wild species. These results suggest that the oilseed cultivar harbors genes acquired during domestication to control seed shape in this genomic region. This study also provides genetic evidence that domestication arose, at least in part, by selection for the oilseed type from wild species and demonstrates the effectiveness of image-based phenotyping to accelerate discoveries of the genetic basis for small morphological features such as seed size and shape.
For the problem of a low recognition rate and shape feature failure caused by overlapping seedlings and weeds during the development of an intelligent lettuce weeding robot, a method to identify seedling lettuce and weeds based on an image block and support vector machine (SVM) is proposed, which realizes their precise identification and boundary segmentation. The a* channel is used to grayscale the collected image. The Otsu and morphological methods are selected to extract all the green targets in the image. The connected component analysis method is applied to label the green targets with regions of interest (ROIs), and those with pixel areas larger than the area threshold are normalized to 256 × 256 pixels. The image blocking technique is introduced to separately aliquot the normalized ROI, with block sizes of 16 × 16, 32 × 32, and 64 × 64 pixels. On this basis, the image sub-blocks are manually labeled, block by block, to extract three texture features: histogram of oriented gradient (HOG), local binary pattern (LBP), and gray-level co-occurrence matrix (GLCM). With the accuracy of fivefold cross-validation as the optimization objective, a genetic algorithm (GA) is used to optimize the SVM penalty and kernel parameters of 21 groups of research objects (one block size has three texture features, which are arbitrarily combined to form seven research objects, with a total of three block sizes). We compare the recognition performance of the SVM, RF, KNN, and GA-SVM classifiers in a single feature and a combination of fusion strategies through comparative analysis. When the block size is 32 × 32 pixels, the fusion of LBP and GLCM features under the GA-SVM classifier has the highest accuracy, and the optimal SVM model for the identification of lettuce and weeds in the seedling stage is obtained. For the misidentified image sub-blocks in optimization model recognition, an image block reconstruction method based on the comparison of the center point and eight-neighbor label value is proposed, and this is combined with the proportion of image blocks of two labels for comprehensive judgment. The center point label value is reconstructed to the improve recognition accuracy. Experimental results show that the average precision, recall, and F1 score of the proposed method are 0.9473, 0.9529, and 0.9498, respectively, and those of images without overlapping leaves can all reach 1, thus providing a theoretical basis for crop recognition and segmentation.
The plant factory is a form of controlled environment agriculture (CEA) which is offers a promising solution to the problem of food security worldwide. Plant growth parameters need to be acquired for process control and yield estimation in plant factories. In this paper, we propose a fast and non-destructive framework for extracting growth parameters. Firstly, ToF camera (Microsoft Kinect V2) is used to obtain the point cloud from the top view, and then the lettuce point cloud is separated. According to the growth characteristics of lettuce, a geometric method is proposed to complete the incomplete lettuce point cloud. The treated point cloud has a high linear correlation with the actual plant height ( R 2 = 0.961), leaf area ( R 2 = 0.964), and fresh weight ( R 2 = 0.911) with a significant improvement compared to untreated point cloud. The result suggests our proposed point cloud completion method have has the potential to tackle the problem of obtaining the plant growth parameters from a single 3D view with occlusion.
Lettuce grown in indoor farms under fully artificial light is susceptible to a physiological disorder known as tip-burn. A vital factor that controls plant growth in indoor farms is the ability to adjust the growing environment to promote faster crop growth. However, this rapid growth process exacerbates the tip-burn problem, especially for lettuce. This paper presents an automated detection of tip-burn lettuce grown indoors using a deep-learning algorithm based on a one-stage object detector. The tip-burn lettuce images were captured under various light and indoor background conditions (under white, red, and blue LEDs). After augmentation, a total of 2333 images were generated and used for training using three different one-stage detectors, namely, CenterNet, YOLOv4, and YOLOv5. In the training dataset, all the models exhibited a mean average precision (mAP) greater than 80% except for YOLOv4. The most accurate model for detecting tip-burns was YOLOv5, which had the highest mAP of 82.8%. The performance of the trained models was also evaluated on the images taken under different indoor farm light settings, including white, red, and blue LEDs. Again, YOLOv5 was significantly better than CenterNet and YOLOv4. Therefore, detecting tip-burn on lettuce grown in indoor farms under different lighting conditions can be recognized by using deep-learning algorithms with a reliable overall accuracy. Early detection of tip-burn can help growers readjust the lighting and controlled environment parameters to increase the freshness of lettuce grown in plant factories.
MAIN CONCLUSION: The combination of image-based phenotyping with in-depth anatomical analysis allows for a thorough investigation of plant physiological plasticity in acclimation, which is driven by environmental conditions and mediated by anatomical traits. Understanding the ability of plants to respond to fluctuations in environmental conditions is critical to addressing climate change and unlocking the agricultural potential of crops both indoor and in the field. Recent studies have revealed that the degree of eco-physiological acclimation depends on leaf anatomical traits, which show stress-induced alterations during organogenesis. Indeed, it is still a matter of debate whether plant anatomy is the bottleneck for optimal plant physiology or vice versa. Here, we cultivated 'Salanova' lettuces in a phenotyping chamber under two different vapor pressure deficits (VPDs; low, high) and watering levels (well-watered, low-watered); then, plants underwent short-term changes in VPD. We aimed to combine high-throughput phenotyping with leaf anatomical analysis to evaluate their capability in detecting the early stress signals in lettuces and to highlight the different degrees of plants' eco-physiological acclimation to the change in VPD, as influenced by anatomical traits. The results demonstrate that well-watered plants under low VPD developed a morpho-anatomical structure in terms of mesophyll organization, stomatal and vein density, which more efficiently guided the acclimation to sudden changes in environmental conditions and which was not detected by image-based phenotyping alone. Therefore, we emphasized the need to complement high-throughput phenotyping with anatomical trait analysis to unveil crop acclimation mechanisms and predict possible physiological behaviors after sudden environmental fluctuations due to climate changes.
The collection and analysis of large amounts of information on a plant-by-plant basis contributes to the development of precision fertigation and may be achieved by combining remote-sensing technology with high-throughput phenotyping methods. Here, lettuce plants (Lactuca sativa) were grown under optimal and suboptimal nitrogen and irrigation treatments from seedlings to harvest. A Plantarray system was used to calculate and log weights, daily transpiration, and momentary transpiration rates throughout the experiment. From 15 d after planting until experiment termination, the entire array of plants was imaged hourly (from 09.00 h to 14.00 h) using a hyperspectral moving camera. Three vegetation indices were calculated from the plants' reflectance signal: red-edge chlorophyll index (RECI), photochemical reflectance index (PRI), and water index (WI), and combined treatments, physiological measurements, and vegetation indices were compared. RECI values differed significantly between nitrogen treatments from the first day of imaging, and WI values distinguished well-irrigated from drought-treated groups before detecting significant differences in daily transpiration rate. The PRI, calculated hourly during the drought-treatment phase, changed with the momentary transpiration rate. Thus, hyperspectral imaging might be used in growing facilities to detect nitrogen or water shortages in plants before their physiological response affects yields.
Growth traits, such as fresh weight, diameter, and leaf area, are pivotal indicators of growth status and the basis for the quality evaluation of lettuce. The time-consuming, laborious and inefficient method of manually measuring the traits of lettuce is still the mainstream. In this study, a three-stage multi-branch self-correcting trait estimation network (TMSCNet) for RGB and depth images of lettuce was proposed. The TMSCNet consisted of five models, of which two master models were used to preliminarily estimate the fresh weight (FW), dry weight (DW), height (H), diameter (D), and leaf area (LA) of lettuce, and three auxiliary models realized the automatic correction of the preliminary estimation results. To compare the performance, typical convolutional neural networks (CNNs) widely adopted in botany research were used. The results showed that the estimated values of the TMSCNet fitted the measurements well, with coefficient of determination ( R 2 ) values of 0.9514, 0.9696, 0.9129, 0.8481, and 0.9495, normalized root mean square error (NRMSE) values of 15.63, 11.80, 11.40, 10.18, and 14.65% and normalized mean squared error (NMSE) value of 0.0826, which was superior to compared methods. Compared with previous studies on the estimation of lettuce traits, the performance of the TMSCNet was still better. The proposed method not only fully considered the correlation between different traits and designed a novel self-correcting structure based on this but also studied more lettuce traits than previous studies. The results indicated that the TMSCNet is an effective method to estimate the lettuce traits and will be extended to the high-throughput situation. Code is available at https://github.com/lxsfight/TMSCNet.git.
Reproduction assets foundThe paper uses the public Autonomous Greenhouses Challenge 3 dataset (RGB/depth lettuce images with FW/DW/H/D/LA measurements) and states author code availability on GitHub.Code · publicCode is available at https://github.com/lxsfight/TMSCNet.git .Open asset ↗github.com/lxsfight/TMSCNetlines:1-41Code / dataset availability confirmedEurope PMC · OpenAlex · checked 8 Sept 2026
Fresh weight is a widely used growth indicator for quantifying crop growth. Traditional fresh weight measurement methods are time-consuming, laborious, and destructive. Non-destructive measurement of crop fresh weight is urgently needed in plant factories with high environment controllability. In this study, we proposed a multi-modal fusion based deep learning model for automatic estimation of lettuce shoot fresh weight by utilizing RGB-D images. The model combined geometric traits from empirical feature extraction and deep neural features from CNN. A lettuce leaf segmentation network based on U-Net was trained for extracting leaf boundary and geometric traits. A multi-branch regression network was performed to estimate fresh weight by fusing color, depth, and geometric features. The leaf segmentation model reported a reliable performance with a mIoU of 0.982 and an accuracy of 0.998. A total of 10 geometric traits were defined to describe the structure of the lettuce canopy from segmented images. The fresh weight estimation results showed that the proposed multi-modal fusion model significantly improved the accuracy of lettuce shoot fresh weight in different growth periods compared with baseline models. The model yielded a root mean square error (RMSE) of 25.3 g and a coefficient of determination ( R 2 ) of 0.938 over the entire lettuce growth period. The experiment results demonstrated that the multi-modal fusion method could improve the fresh weight estimation performance by leveraging the advantages of empirical geometric traits and deep neural features simultaneously.
Reproduction assets foundThe paper's phenotyping inputs (top-view RGB and aligned depth images of 388 lettuces with destructively measured traits) come from the publicly available 3rd Autonomous Greenhouse Challenge Online Challenge Lettuce Images dataset, with an explicit public URL in the data availability statement. No author analysis code,Dataset · publicPublicly available datasets were analyzed in this study. This data can be found here: https://data.4tu.nl/articles/dataset/3rd_Autonomous_Greenhouse_Challenge_Online_Challenge_Lettuce_Images/15023088 .Open asset ↗data.4tu.nl · 15023088lines:657-691Plant phenotyping relevance matchEurope PMC · checked 15 Sept 2026
Guo Z, Li B, Du J, Shen F, Zhao Y, Deng Y, Kuang Z, Tao Y, Wan M, Lu X, Wang D, Wang Y, Han Y, Wei J, Li L, Guo X, Zhao C, Yang X.
LettuceVisualization / data management
As a globally popular leafy vegetable and a representative plant of the Asteraceae family, lettuce has great economic and academic significance. In the last decade, high-throughput sequencing, phenotyping, and other multi-omics data in lettuce have accumulated on a large scale, thus increasing the demand for an integrative lettuce database. Here, we report the establishment of a comprehensive lettuce database, LettuceGDB (https://www.lettucegdb.com/). As an omics data hub, the current LettuceGDB includes two reference genomes with detailed annotations; re-sequencing data from over 1000 lettuce varieties; a collection of more than 1300 worldwide germplasms and millions of accompanying phenotypic records obtained with manual and cutting-edge phenomics technologies; re-analyses of 256 RNA sequencing datasets; a complete miRNAome; extensive metabolite information for representative varieties and wild relatives; epigenetic data on the genome-wide chromatin accessibility landscape; and various lettuce research papers published in the last decade. Five hierarchically accessible functions (Genome, Genotype, Germplasm, Phenotype, and O-Omics) have been developed with a user-friendly interface to enable convenient data access. Eight built-in tools (Assembly Converter, Search Gene, BLAST, JBrowse, Primer Design, Gene Annotation, Tissue Expression, Literature, and Data) are available for data downloading and browsing, functional gene exploration, and experimental practice. A community forum is also available for information sharing, and a summary of current research progress on different aspects of lettuce is included. We believe that LettuceGDB can be a comprehensive functional database amenable to data mining and database-driven exploration, useful for both scientific research and lettuce breeding.
Growth indices can quantify crop productivity and establish optimal environmental, nutritional, and irrigation control strategies. A convolutional neural network (CNN)-based model is presented for estimating various growth indices (i.e., fresh weight, dry weight, height, leaf area, and diameter) of four varieties of greenhouse lettuce using red, green, blue, and depth (RGB-D) data obtained using a stereo camera. Data from an online autonomous greenhouse challenge (Wageningen University, June 2021) were employed in this study. The data were collected using an Intel RealSense D415 camera. The developed model has a two-stage CNN architecture based on ResNet50V2 layers. The developed model provided coefficients of determination from 0.88 to 0.95, with normalized root mean square errors of 6.09%, 6.30%, 7.65%, 7.92%, and 5.62% for fresh weight, dry weight, height, diameter, and leaf area, respectively, on unknown lettuce images. Using red, green, blue (RGB) and depth data employed in the CNN improved the determination accuracy for all five lettuce growth indices due to the ability of the stereo camera to extract height information on lettuce. The average time for processing each lettuce image using the developed CNN model run on a Jetson SUB mini-PC with a Jetson Xavier NX was 0.83 s, indicating the potential for the model in fast real-time sensing of lettuce growth indices.
Reproduction assets foundThe paper's phenotyping inputs (388 RGB-D lettuce image pairs with destructive growth-index measurements from the Third Autonomous Greenhouse Challenge) are a third-party public dataset explicitly stated to be publicly available at 4TU.ResearchData, with the DOI 10.4121/15023088.v1 cited in the text and figure captionsDataset · publicThe dataset is available in online: https://doi.org/10.4121/15023088.v1 [ 30 ].Open asset ↗10.4121/15023088.v1lines:518-697Plant phenotyping relevance matchEurope PMC · checked 8 Sept 2026
To solve the problem of low survival rate caused by unscreened transplanting of seedlings. This study proposed a selective transplanting method of leafy vegetable seedlings based on the ResNet 18 network. Lettuce seedlings were selected as the research object, and a total of 3,388 images were obtained in the dataset. The images were randomly divided into the training set, validation set, and test set in the ratio of 6:2:2. The ResNet 18 network was used to perform transfer learning after tuning, identifying, and classifying leafy vegetable seedlings, and then establishing a model to screen leafy vegetable seedlings. The results showed that the optimal detection accuracy of the presence and health of seedlings in the training data set was above 100%, and the model loss remained at around 0.005. Nine hundred seedlings were selected for the validation test, and the screening accuracy rate was 97.44%, the precision rate of healthy seedlings was 97.56%, the recall rate was 97.34%, the precision rate of unhealthy seedlings was 92%, and the recall rate was 92.62%, which was better than the screening model based on the physical characteristics of seedlings. If they were identified as unhealthy seedlings, the manipulator would remove them during the transplanting process and perform the seedling replenishment operation to increase the survival rate of the transplanted seedlings. Moreover, the seedling image is extracted by background removal technology, so the model processing time for a single image is only 0.0129 s. This research will provide technical support for the selective transplantation of leafy vegetable seedlings.
Real-time, continuous young seedling-growth measurement improves plant factory stabilization and productivity. Projected leaf area (PLA) based on seedling top-view images is a useful growth index, and easy to measure continuously for large seedling populations. However, it is difficult to automatically determine PLA with a high degree of accuracy, because RGB image color-balance fluctuates with plant growth, leaf movement, and environment. Therefore, we developed a technique for determining PLA on nursery-grown lettuce seedlings. Using a Raspberry Pi 3 microcomputer with a camera module placed above the seedlings, RGB images of 153 seedlings were obtained every 20 min from day 6 to 15 after sowing. Seedling PLA images were obtained by binarization and separation of the leaves from the background. To assess binarization accuracy, we used an Intersection over Union (IoU) index to compare the standard Excess Green (ExG) method, an optimized ExG method (O-ExG), and the artificial neural network U-Net method. Results showed that O-ExG was optimal under the experimental conditions tested. PLA and circadian rhythm amplitude extracted from PLA-image time-series data were independent, implying they can be used together for growth prediction. These findings improve the accuracy of imagebased growth prediction and have practical application in plant factories.
The currently available methods for evaluating most biochemical traits of plant phenotyping are destructive and have extremely low throughput. However, hyperspectral techniques can non-destructively obtain the spectral reflectance characteristics of plants, which can provide abundant biophysical and biochemical information. Therefore, plant spectra combined with machine learning algorithms can be used to predict plant phenotyping traits. However, the raw spectral reflectance characteristics contain noise and redundant information, thus can easily affect the robustness of the models developed via multivariate analysis methods. In this study, two end-to-end deep learning models were developed based on 2D convolutional neural networks (2DCNN) and fully connected neural networks (FCNN; Deep2D and DeepFC, respectively) to rapidly and non-destructively predict the phenotyping traits of lettuces from spectral reflectance. Three linear and two nonlinear multivariate analysis methods were used to develop models to weigh the performance of the deep learning models. The models based on multivariate analysis methods require a series of manual feature extractions, such as pretreatment and wavelength selection, while the proposed models can automatically extract the features in relation to phenotyping traits. A visible near-infrared hyperspectral camera was used to image lettuce plants growing in the field, and the spectra extracted from the images were used to train the network. The proposed models achieved good performance with a determination coefficient of prediction ( Rp2 ) of 0.9030 and 0.8490 using Deep2D for soluble solids content and DeepFC for pH, respectively. The performance of the deep learning models was compared with five multivariate analysis method. The quantitative analysis showed that the deep learning models had higher Rp2 than all the multivariate analysis methods, indicating better performance. Also, wavelength selection and different pretreatment methods had different effects on different multivariate analysis methods, and the selection of appropriate multivariate analysis methods and pretreatment methods increased more time and computational cost. Unlike multivariate analysis methods, the proposed deep learning models did not require any pretreatment or dimensionality reduction and thus are more suitable for application in high-throughput plant phenotyping platforms. These results indicate that the deep learning models can better predict phenotyping traits of plants using spectral reflectance.
Abstract In recent years, 3D geometry has become increasingly important for plant phenotyping. The purpose of this study is to assess the measuring accuracy of a photogrammetric method based on SfM-MVS using a series of images captured by smartphone. Butterhead lettuce and the rubber plant were the two different plants used in this study. For each single plant, the images were captured from multiple views. A photogrammetry software took image input and converted them into 3D point cloud. Finally, the plant height was computed from the point cloud. Comparing the computed values to the actual values, the RMSE of the plant height was 0.28 and 0.43 for butterhead lettuce and the rubber plant, respectively. A correlation of R 2 ≥ 0.94 to the reference measurement demonstrated that the photogrammetric approach is well suited for evaluation of the plant. The proposed method is simple and cost-effective by using a readily accessible device and software to reconstruct a point cloud model.
Abstract Wild species of lettuce ( Lactuca sp.) are thought to have first been domesticated for oilseed contents to provide seed oil for human consumption. Although seed morphology is an important trait contributing to oilseed in lettuce, the underlying genetic mechanisms remain elusive. Since lettuce seeds are small, a manual phenotypic determination required for a genetic dissection of such traits is challenging. In this study, we built and applied an instance segmentation-based seed morphology quantification pipeline to measure traits in seeds generated from a cross between the domesticated oilseed type cultivar ‘Oilseed’ and the wild species ‘UenoyamaMaruba’ in an automated manner. Quantitative trait locus (QTL) mapping following ddRAD-seq revealed 11 QTLs linked to 7 seed traits (area, width, length, length-to-width ratio, eccentricity, perimeter length, and circularity). Remarkably, the three QTLs with the highest LOD scores, qLWR-3 . 1, qECC-3 . 1 , and qCIR-3 . 1 , for length-to-width ratio, eccentricity, and circularity, respectively, mapped to linkage group 3 (LG3) around 161.5 to 214.6 Mb, a region previously reported to be associated with domestication traits from wild species. These results suggest that the oilseed cultivar harbors genes acquired during domestication to control seed shape in this genomic region. This study also provides genetic evidence that domestication arose, at least in part, by selection for the oilseed type from wild species and demonstrates the effectiveness of image-based phenotyping to accelerate discoveries of the genetic basis for small morphological features such as seed size and shape.
Recent years have seen an increased effort in the detection of plant stresses and diseases using non-invasive sensors and deep learning methods. Nonetheless, no studies have been made on dense plant canopies, due to the difficulty in automatically zooming into each plant, especially in outdoor conditions. Zooming in and zooming out is necessary to focus on the plant stress and to precisely localize the stress within the canopy, for further analysis and intervention. This work concentrates on tip-burn, which is a plant stress affecting lettuce grown in controlled environmental conditions, such as in plant factories. We present a new method for tip-burn stress detection and localization, combining both classification and self-supervised segmentation to detect, localize, and closely segment the stressed regions. Starting with images of a dense canopy collecting about 1,000 plants, the proposed method is able to zoom into the tip-burn region of a single plant, covering less than 1/10th of the plant itself. The method is crucial for solving the manual phenotyping that is required in plant factories. The precise localization of the stress within the plant, of the plant within the tray, and of the tray within the table canopy allows to automatically deliver statistics and causal annotations. We have tested our method on different data sets, which do not provide any ground truth segmentation mask, neither for the leaves nor for the stresses; therefore, the results on the self-supervised segmentation is even more impressive. Results show that the accuracy for both classification and self supervised segmentation is new and efficacious. Finally, the data set used for training test and validation is currently available on demand.
BACKGROUND: Classification and phenotype identification of lettuce leaves urgently require fine quantification of their multi-semantic traits. Different components of lettuce leaves undertake specific physiological functions and can be quantitatively described and interpreted using their observable properties. In particular, petiole and veins determine mechanical support and material transport performance of leaves, while other components may be closely related to photosynthesis. Currently, lettuce leaf phenotyping does not accurately differentiate leaf components, and there is no comparative evaluation for positive-back of the same lettuce leaf. In addition, a few traits of leaf components can be measured manually, but it is time-consuming, laborious, and inaccurate. Although several studies have been on image-based phenotyping of leaves, there is still a lack of robust methods to extract and validate multi-semantic traits of large-scale lettuce leaves automatically. RESULTS: In this study, we developed an automated phenotyping pipeline to recognize the components of detached lettuce leaves and calculate multi-semantic traits for phenotype identification. Six semantic segmentation models were constructed to extract leaf components from visible images of lettuce leaves. And then, the leaf normalization technique was used to rotate and scale different leaf sizes to the "size-free" space for consistent leaf phenotyping. A novel lamina-based approach was also utilized to determine the petiole, first-order vein, and second-order veins. The proposed pipeline contributed 30 geometry-, 20 venation-, and 216 color-based traits to characterize each lettuce leaf. Eleven manually measured traits were evaluated and demonstrated high correlations with computation results. Further, positive-back images of leaves were used to verify the accuracy of the proposed method and evaluate the trait differences. CONCLUSIONS: The proposed method lays an effective strategy for quantitative analysis of detached lettuce leaves' fine structure and components. Geometry, color, and vein traits of lettuce leaf and its components can be comprehensively utilized for phenotype identification and breeding of lettuce. This study provides valuable perspectives for developing automated high-throughput phenotyping application of lettuce leaves and the improvement of agronomic traits such as effective photosynthetic area and vein configuration.
Plant breeders, scientists, and commercial producers commonly use growth rate as an integrated signal of crop productivity and stress. Plant growth monitoring is often done destructively via growth rate estimation by harvesting plants at different growth stages and simply weighing each individual plant. Within plant breeding and research applications, and more recently in commercial applications, non-destructive growth monitoring is done using computer vision to segment plants in images from the background, either in 2D or 3D, and relating these image-based features to destructive biomass measurements. Recent advancements in machine learning have improved image-based localization and detection of plants, but such techniques are not well suited to make biomass predictions when there is significant self-occlusion or occlusion from neighboring plants, such as those encountered under leafy green production in controlled environment agriculture. To enable prediction of plant biomass under occluded growing conditions, we develop an end-to-end deep learning approach that directly predicts lettuce plant biomass from color and depth image data as provided by a low cost and commercially available sensor. We test the performance of the proposed deep neural network for lettuce production, observing a mean prediction error of 7.3% on a comprehensive test dataset of 864 individuals and substantially outperforming previous work on plant biomass estimation. The modeling approach is robust to the busy and occluded scenes often found in commercial leafy green production and requires only measured mass values for training. We then demonstrate that this level of prediction accuracy allows for rapid, non-destructive detection of changes in biomass accumulation due to experimentally induced stress induction in as little as 2 days. Using this method growers may observe and react to changes in plant-environment interactions in near real time. Moreover, we expect that such a sensitive technique for non-destructive biomass estimation will enable novel research and breeding of improved productivity and yield in response to stress.
Reproduction assets foundThe article provides an authors' public GitHub repository containing the analysis code for the RGB-D deep learning biomass estimation pipeline. The raw image/biomass dataset is only available on request (no public deposit).Code · publicCode available at https://github.com/NicoBux/Plant-Biomass-Monitoring .Open asset ↗NicoBux/Plant-Biomass-Monitoringlines:466-524Plant phenotyping relevance matchOpenAlex · Crossref · Europe PMC · checked 8 Sept 2026
Fluorescence imaging has shown great potential in non-invasive plant monitoring and analysis. However, current systems have several limitations, such as bulky size, high cost, contact measurement, and lack of multifunctionality, which may hinder its applications in a wide range of settings including indoor vertical farming. Herein, we developed a compact handheld fluorescence imager enabling multipurpose plant phenotyping, such as continuous photosynthetic activity monitoring and non-destructive anthocyanin quantification. The compact imager comprises of pulse-amplitude-modulated multi-color light emitting diodes (LEDs), optimized light illumination and collection, dedicated driver circuit board, miniaturized charge-coupled device camera, and associated image analytics. Experiments conducted in drought stressed lettuce proved that the novel imager could quantitatively evaluate the plant stress by the non-invasive measurement of photosynthetic activity efficiency. Moreover, a non-invasive and fast quantification of anthocyanins in green and red Batavia lettuce leaves had excellent correlation (>84%) with conventional destructive biochemical analysis. Preliminary experimental results emphasize the high throughput monitoring capability and multifunctionality of our novel handheld fluorescence imager, indicating its tremendous potential in modern agriculture.
Water availability is a major constraint for crop production worldwide. Remote sensing provides an ideal mean to monitor vegetation status from the canopy to the ecosystem scale. Classical approaches have mainly used the reduced vegetation development as a stress indicator. This research discusses short-term reactions on a plant to drought stress as well as their corresponding effects on different hyperspectral remote sensing metrics. As a first effect, a reduction in the plant water content results in a drop in the leaf turgor, which changes the leaf orientation. This effect changes the canopy structure, changing the near-infrared reflectance. Second, a water shortage in a plant induces stomatal closure, which limits the gas exchange. This reduces the amount of CO 2 that the photosynthetic apparatus can assimilate, causing an imbalance between the energy demanded by the CO 2 assimilation part and the energy provided by the photosynthetic light reactions. As a consequence, an alternative electron sink is needed at the light reactions side. This is provided for by a series of mechanisms collectively known as non-photochemical quenching (NPQ). The increase in NPQ leads to a change in the hyperspectral photochemical index (PRI) and to a change in the sun-induced chlorophyll fluorescence (SIF) emission. The latter consists of the radiation that is re-emitted by a chlorophyll molecule. To evaluate the effect of a drought stress on these remote sensing metrics, the hyperspectral reflectance and the SIF emission were measured over a mustard and a lettuce canopy. At the same time, the soil moisture and weather conditions were monitored. The PRI shows a clear diurnal pattern, in which the PRI is anticorrelated with the photosynthetically active radiation (PAR). The pattern is more expressed for stressed days. The canopy structure’s reaction to drought stress is very species-specific, as this reaction is affected by the presence of woody material in the canopy. The SIF reaction only becomes clear after it has been normalized for the PAR and for the canopy structure. The link between SIF and PAR depends on the plant stress status. We argue that the combination of these three factors (PRI, SIF and reflectance) provide solid information on the degree of water limitation in the plant.
Nowadays, about 50% of the global yield loss is due to climate change. Increasing Vapor Pressure Deficit (VPD) and drought are among the principal environmental stressors, affecting stomatal regulation and reducing plant photosynthesis and biomass accumulation. Recent studies have revealed that the extent of plant acclimation is closely related to the anatomical traits of the leaves, which change with environmental conditions. It is not yet clear how the interaction between these environmental factors affects plant morpho-physiological development and plant capacity of acclimation under changing conditions. To fill this gap, in this study we used a high-throughput phenotyping facility (at the IPK-Gatersleben, Germany) to grow two lettuce cultivars ( Lactuca sativa L. var. capitata ) with green and red leaves under different VPDs (low and high) and watering regimes (well-watered, WW, and low watered, LW regimes). Two trials were performed: the first trial was conducted at a VPD of 0.7 kPa (low VPD) and the second at 1.4 kPa (high VPD), both with WW and WD conditions. After 12 days of cultivation in the phenotyping chamber, the environmental conditions were switched, and plants were kept for 5 days at the opposite VPD to evaluate their acclimation ability. RGB imaging was applied to track changes in morphological parameters, near-infrared camera (NIR) was used to estimate plant-water relationships, and FLUO made it possible to evaluate changes in photosystem II reflecting optimal/stressful conditions. At the end of the experimental trials, the leaf samples were characterized in terms of stomatal and mesophyll traits by light microscopy. A specific focus was dedicated to exploring how stomata regulation and water use efficiency affect carbon gain and biomass allocation in pre-acclimated lettuces to different environmental conditions (VPDs) and hence undergoing sudden changes in the VPD. To test the influence of the different independent factors: i) VPD, ii) cultivar (C), iii) water (W) on the dependent variables, a three-way analysis of variance (ANOVA) was performed. Additionally, correlation plots and the principal component analysis were performed to explore correlations between morpho-anatomical and phenotypic data points. The results showed that WW plants at low VPD developed a morpho-anatomical structure in terms of mesophyll organization, stomatal and vein density which more efficiently guided acclimation to sudden changes in the environmental conditions and which was not detected by image-based phenotyping alone. Therefore, we emphasized the need to complement high-throughput phenotyping with the analysis of anatomical traits to unravel the mechanisms of crop acclimation under sudden fluctuation in environmental conditions due to climate change. Such an approach can help improving knowledge on how stomatal regulation and carbon allocation affect productivity in warmer areas and drier climates, with high impact also for the design of cultivation protocols for sustainable indoor farming.
In response to the challenges in providing real-time extraction of crop biophysical signatures, computer vision in computational crop phenotyping highlights the opportunities of computational intelligence solutions. Shadow and angular brightness due to the presence of photosynthetic light unevenly illuminate crop canopy. In this study, a novel vegetation index named artificial bee colony-optimized visible band oblique dipyramid greenness index (vODGIabc) was proposed to enhance vegetation pixels by correcting the saturation and brightness levels, and the ratio of visible RGB reflectance intensities. Consumer-grade smartphone was used to acquire indoor and outdoor aquaponic lettuce images daily for full 6-week crop life cycle. The introduced saturation rectification coefficient (Ω), value rectification coefficient (ν), green–red wavelength adjustment factor (α), and green–blue wavelength adjustment factor (β) on the original triangular greenness index resulted in 3D canopy reflectance spectrum with two oblique tetrahedrons formed by connecting the vertices of visible RGB band reflectance and maximum wavelength point map to corresponding saturation and value of lettuce-captured images. Hybrid neighborhood component analysis (NCA), minimum redundancy maximum relevance (MRMR), Pearson’s correlation coefficient (PCC), and analysis of variance (ANOVA) weighted most of the canopy area, energy, and homogeneity. Strong linear relationships were exhibited by using vODGIabc in estimating lettuce crop fresh weight, height, number of spanning leaves, leaf area index, and growth stage with R2 values of 0.936 8 for InceptionV3, 0.957 4 for ResNet101, 0.961 2 for ResNet101, 0.999 9 for Gaussian processing regression, and accuracy of 88.89% for ResNet101, respectively. This low-cost approach on developing greenness index for biophysical signatures estimation proved to be more accurate than the previously established triangular greenness index (TGI) using RGB smartphone camera.
In order to meet the needs of dynamic continuous monitoring of soil-plant-atmosphere continuum (SPAC), a new soil, plant, atmosphere analysis system has been established based on an intelligent weighing system (IWS). Four types of irrigation treatments (90%, 80%, 70%, and 60% of field capacity (FC)) were conducted on lettuce (Lactuca sativa var. ramosa Hort.) for two-season planting experiments. Regarding the soil, the relative system weight of IWS showed a significant linear correlation with the soil volumetric moisture content (SWC) (R2 = 0.64–0.94). When the SWC increased by 1.00%, the soil weight increased by 56–62 g. Regarding plants, the IWS also clearly reflected the changes in plant weight gain, transpiration rate, and stomatal conductance at different growth stages. After verification, the relative errors of the transpiration rate and stomatal conductance measured by the IWS were −9.60–22.30% and −7.20–22.20%, respectively. Regarding the atmospheric environment, the variation trend of the crop evapotranspiration (ETc) based on the IWS and the reference crop evapotranspiration (ET0) calculated with meteorological parameters were consistent. However, the numerical difference was in the uncertainty of the crop coefficient (Kc). The ETc of lettuce under the 80% FC treatment was the highest. Accordingly, a daily online measurement method for Kc was established. The Kc values of lettuce at different growth stages were 0.88, 1.22, and 2.43, respectively. The growth, yield, and water use efficiency (WUE) of crops under 80% FC treatment compared with other treatments significantly increased by 11.07–21.05%, 0.91–9.89%, and 2.16–15.80%, respectively. Therefore, the 80% FC was adopted as the irrigation low limit of potted lettuce. The experimental results provide a theoretical basis for further guiding crop irrigation.
Focusing on non-destructive and automated acquisition of plant phenotypic parameters,this extended abstract proposed an end-to-end deep RNN based network structure for single perspective sparse raw point cloud regression task called DRN. It has been proven to achieve accuracy improvements in PointNet++ and PonitCNN when it comes to regression of lettuce plant height. We believe DRN structure is suitable for feature extraction from plant point cloud data and regression of spatial distance related plant phenotypes like plant height.
Urban agriculture can be shortly defined as the growing of plants and/or the livestock husbandry in and around cities. Although it has been a common occupation for the urban population all along, recently there is a growing interest in it both from public bodies and researchers, as well as from ordinary citizens who want to engage in self-cultivation. The modern citizen, though, will hardly find the free time to grow his own vegetables as it is a process that requires, in addition to knowledge and disposition, consistency. Given the above considerations, the purpose of this work was to develop an economic robotic system for the automatic monitoring and management of an urban garden. The robotic system was designed and built entirely from scratch. It had to have suitable dimensions so that it could be placed in a balcony or a terrace, and be able to scout vegetables from planting to harvest and primarily conduct precision irrigation based on the growth stage of each plant. Fertigation and weed control will also follow. For its development, a number of technologies were combined, such as Cartesian robots' motion, machine vision, deep learning for the identification and detection of plants, irrigation dosage and scheduling based on plants' growth stage, and cloud storage. The complete process of software and hardware development to a robust robotic platform is described in detail in the respective sections. The experimental procedure was performed for lettuce plants, with the robotic system providing precise movement of its actuator and applying precision irrigation based on the specific needs of the plants.
In the aquaponic system, plant nutrients bioavailable from fish excreta are not sufficient for optimal plant growth. Accurate and timely monitoring of the plant’s nutrient status grown in aquaponics is a challenge in order to maintain the balance and sustainability of the system. This study aimed to integrate color imaging and deep convolutional neural networks (DCNNs) to diagnose the nutrient status of lettuce grown in aquaponics. Our approach consists of multi-stage procedures, including plant object detection and classification of nutrient deficiency. The robustness and diagnostic capability of proposed approaches were evaluated using a total number of 3000 lettuce images that were classified into four nutritional classes—namely, full nutrition (FN), nitrogen deficiency (N), phosphorous deficiency (P), and potassium deficiency (K). The performance of the DCNNs was compared with traditional machine learning (ML) algorithms (i.e., Simple thresholding, K-means, support vector machine; SVM, k-nearest neighbor; KNN, and decision Tree; DT). The results demonstrated that the deep proposed segmentation model obtained an accuracy of 99.1%. Also, the deep proposed classification model achieved the highest accuracy of 96.5%. These results indicate that deep learning models, combined with color imaging, provide a promising approach to timely monitor nutrient status of the plants grown in aquaponics, which allows for taking preventive measures and mitigating economic and production losses. These approaches can be integrated into embedded devices to control nutrient cycles in aquaponics.
Plant factories with artificial lighting (PFALs), with well-insulated and airtight structures, enable the production of large quantities of high-quality plants year-round while achieving high resource use efficiency. However, despite the controlled environment in PFALs, variations in plant individuals have been found, which affect productivity in PFAL operations. Plant phenotyping plays a crucial role in understanding how the surrounding microenvironment affects variations in plant phenotypes. In the current study, a modular phenotyping system for seedling production was developed, focusing on practicality and scalability in commercial PFALs. Experiments on seedlings, which strongly affect productivity, were conducted to obtain cotyledon unfolding time and the time series projected area of cotyledons and true leaves of individual seedlings of romaine lettuce (Lactuca sativa L. var. longifolia), using RGB images. This was also undertaken to analyze how the surrounding microenvironment of photosynthetic photon flux densities and nutrients affect growth variations for plant cohort research. In agreement with the actual measurements, variations in seedling growth were identified even under similar microenvironments. Furthermore, the results demonstrated larger variations in seedlings with higher relative growth. Aiming for simplified interactions of phenotypes with the microenvironment, management, and genotype, seedling selection and breeding with plant production in PFALs may enable plant uniformity and higher productivity.
This article applies deep learning and electromechanical technology to plant phenotype measurement. First, an electromechanical device is designed to collect plant phenotype images, which solves the difficulty of collecting deep learning training data. The data set required for deep learning model training for plant phenotype detection is made by an automated method. This paper takes the Lactuca sativa plant image as an example and uses the ASM‐based data enhancement method to solve the problem of insufficient image data of Lactuca sativa leaf pests and effectively avoid the phenomenon of overfitting. The plant image recognition method based on deep learning proposed breaks through the limitations of plant local feature recognition, gets rid of the limitation of highly specialized data collection, lowers the threshold of plant image recognition, and has advantages in recognition speed and accuracy. This method requires a large amount of training data. In the future, we can explore the collection of massive plant pictures from the Internet as a training set to achieve rapid iteration and optimization of the model.
Computer vision systems’ interest in food grading has been increasing and adopted due to the non-destructive and contactless features of the process. Aquaponics technique, on the other hand, is a farming method that combines a recirculating aquaculture system and soilless hydroponics agriculture promising to be one of the answers to sustainability in the food industry. Lack of intelligent real-time approaches to monitor and track plant growth is hindering the transition of aquaponic systems towards automation and commercialization. Computer vision can promote further contributions in smart applications in aquaponics; therefore, a methodology is proposed to measure in real-time the growth rate and fresh weight of crops in multi-instance setups. The proposed system uses image-processing techniques, deep learning, and regression analysis to estimate the size of the crops as they grow using image segmentation. Then, a correlation between the size of the crops and their fresh weight is modelled. For common little gem romaine lettuce, the size of crops and fresh weight is estimated with an overall error of 30 mm (18.7%) and 0.5 g (8.3%), respectively.
The identification and control of light stress is the key to the high-yield and high-quality production of leafy vegetables in a controlled environment. As a widely grown vegetable in the plant factory, lettuce is responsive to light intensity. Strong or weak light will seriously affect its yield and quality. These differences can be articulated by images and then classified by fine-grained classification methods. Deep learning is commonly used in crop image classification due to its fast and convenient advantages, but conventional fine-grained recognition approaches are still extremely challenging in dealing with this kind of inter-species classification. To address this issue, this work takes lettuce as the research object and established a set of leaf images for lettuce light stress grading. The leaves were divided into four categories depending on the changes in shoot fresh weight. Then, a hierarchical fusion convolutional neural network architecture (MFC-CNN) based on multi-scale input was constructed to grade the light stress. We firstly separate leaf patches from complete leaf and construct four-scale input to expand local leaf vein and texture information. The multi-scale dataset is fed into the main network at different depths according to the characteristics of CNN in feature extraction. Finally, the network is tested through comparative experiments. The results show that the proposed model has obvious advantages in light stress dataset and generalized dataset.
Focusing on non-destructive and automated acquisition of plant phenotypic parameters,this extended abstract proposed an end-to-end deep RNN based network structure for single perspective sparse raw point cloud regression task called DRN. It has been proven to achieve accuracy improvements in PointNet++ and PonitCNN when it comes to regression of lettuce plant height. We believe DRN structure is suitable for feature extraction from plant point cloud data and regression of spatial distance related plant phenotypes like plant height.
To overcome the challenges related to food security, digital farming has been proposed, wherein the status of a plant using various sensors could be determined in real time. The high-throughput phenotyping platform (HTPP) and analysis with deep learning (DL) are increasingly being used but require a lot of resources. For botanists who have no prior knowledge of DL, the image analysis method is relatively easy to use. Hence, we aimed to explore a pre-trained Arabidopsis DL model to extract the projected area (PA) for lettuce growth pattern analysis. The accuracies of the extract PA of the lettuce cultivar “Nul-chung” with a pre-trained model was measured using the Jaccard Index, and the median value was 0.88 and 0.87 in two environments. Moreover, the growth pattern of green lettuce showed reproducible results in the same environment (p < 0.05). The pre-trained model successfully extracted the time-series PA of lettuce under two lighting conditions (p < 0.05), showing the potential application of a pre-trained DL model of target species in the study of traits in non-target species under various environmental conditions. Botanists and farmers would benefit from fewer challenges when applying up-to-date DL in crop analysis when few resources are available for image analysis of a target crop.
Some plant pigments are strong antioxidants and benefit human health. Given their low cost and popularity, biofortified lettuce cultivars can promote consumption of these compounds. In this study, we assessed anthocyanin, chlorophyll, and carotenoid content in 30 red lettuce genotypes, as well as the Soil Plant Analysis Development index (SPAD) and vegetation indices (CIG, CVI, GNDVI, and NDVI), using high-throughput phenotyping. We calculated vegetation indices by using images of the plant canopy obtained from a remotely piloted aircraft (RPA) together with a red, green, and near-infrared band camera. The SPAD index and leaf pigments were measured through traditional laboratory methods, which are costly and time-consuming. Genetic variability among the genotypes for leaf pigments was confirmed by the Scott-Knott test and UPGMA dendrogram. Anthocyanin was the pigment that most contributed to genetic diversity and highly correlated with CIG, CVI, and GNDVI. These indices also moderately correlated with chlorophylls and carotenoids. Genotypes with higher pigment content, especially anthocyanin, had higher indices. Therefore, the image-based plant phenotyping technique correlated with the traditional methodology and can provide an alternative for indirect selection of plants with high leaf pigment content, especially anthocyanin, in red lettuce.
Plant and plant organ movements are the result of a complex integration of endogenous growth and developmental responses, partially controlled by the circadian clock, and external environmental cues. Monitoring of plant motion is typically done by image-based phenotyping techniques with the aid of computer vision algorithms. Here we present a method to measure leaf movements using a digital inertial measurement unit (IMU) sensor. The lightweight sensor is easily attachable to a leaf or plant organ and records angular traits in real-time for two dimensions (pitch and roll) with high resolution (measured sensor oscillations of 0.36 ± 0.53° for pitch and 0.50 ± 0.65° for roll). We were able to record simple movements such as petiole bending, as well as complex lamina motions, in several crops, ranging from tomato to banana. We also assessed growth responses in terms of lettuce rosette expansion and maize seedling stem movements. The IMU sensors are capable of detecting small changes of nutations (i.e. bending movements) in leaves of different ages and in different plant species. In addition, the sensor system can also monitor stress-induced leaf movements. We observed that unfavorable environmental conditions evoke certain leaf movements, such as drastic epinastic responses, as well as subtle fading of the amplitude of nutations. In summary, the presented digital sensor system enables continuous detection of a variety of leaf motions with high precision, and is a low-cost tool in the field of plant phenotyping, with potential applications in early stress detection.
Chlorophyll fluorescence is interesting for phenotyping applications as it is rich in biological information and can be measured remotely and non-destructively. There are several techniques for measuring and analysing this signal. However, the standard methods use rather extreme conditions, e.g., saturating light and dark adaption, which are difficult to accommodate in the field or in a greenhouse and, hence, limit their use for high-throughput phenotyping. In this article, we use a different approach, extracting plant health information from the dynamics of the chlorophyll fluorescence induced by a weak light excitation and no dark adaption, to classify plants as healthy or unhealthy. To evaluate the method, we scanned over a number of species (lettuce, lemon balm, tomato, basil, and strawberries) exposed to either abiotic stress (drought and salt) or biotic stress factors (root infection using Pythium ultimum and leaf infection using Powdery mildew Podosphaera aphanis). Our conclusions are that, for abiotic stress, the proposed method was very successful, while, for powdery mildew, a method with spatial resolution would be desirable due to the nature of the infection, i.e., point-wise spread. Pythium infection on the roots is not visually detectable in the same way as powdery mildew; however, it affects the whole plant, making the method an interesting option for Pythium detection. However, further research is necessary to determine the limit of infection needed to detect the stress with the proposed method.
This study was carried out to evaluate the feasibility of using Vis/near-infrared (Vis/NIR) spectroscopy for determining the potassium concentration in fresh lettuce leaves and petioles of single-variety lettuce and mixed lettuce leaves of two varieties. Partial least squares (PLS) and radial basis function (RBF) neural network were systemically studied and compared as regressions tools in developing the prediction models. Competitive adaptive reweighted sampling (CARS) variable selection and spectral preprocessing (first- and second-order derivatives) were applied to optimize the performance of predictions. On the basis of these selected optimum wavelengths, the established PLS prediction models provided the coefficients of determination (R²) of 0.83 and 0.71, residual predictive deviations (RPD) were 1.95 and 1.80, and root mean square errors of prediction (RMSEP) were 39.07 and 38.06 mg/100 g for green leaves and petioles, respectively. By comparison, the RBF approach with first-derivative preprocessing spectra was found to provide the best performance of mixed samples, yielding R² of 0.86 and 0.88, RMSEP of 31.20 and 27.63 mg/100 g, and RPD of 2.44 and 2.47 for green leaves and petioles, respectively. The overall results of this study revealed the potential for use of Vis/NIR spectroscopy as an objective and non-destructive method to inspect the potassium concentration of fresh lettuces.
Large-scale screening and assessment of lettuce resources are valuable to help discover significant traits and assist in genetic breeding. In this research, a greenhouse-based vegetable high-throughput phenotyping platform (VHPP) was established to evaluate the multidimensional characteristics of various lettuce varieties. The platform contains an imaging unit with four degrees of freedom (DOFs) to cruise on the crop overhead for image acquisition. The platform also has an automated global semantic phenotyping pipeline (GSPP) for locating pots in sequential images and matching each plant at different growth points. Multidimensional image-based traits were automatically extracted and classified into six categories, including geometry, structure, texture, color, color moment and color indices. We calculated and evaluated 63 static traits (ST) and 189 dynamic traits (DT) by principal component (PC), correlation and heritability analysis, and new PCs provided valuable perspective in describing the lettuce canopy. The results demonstrated the ability of a phenotyping system and pipeline to rapidly investigate and evaluate the growth status of thousands of vegetables. Besides, we identified lots of valuable traits that could be of positive significance in revealing the genetic basis of complex features and exploring the excellent attributes for use in the large-scale lettuce screening and assessment.
Gray mold disease caused by the fungus Botrytis cinerea damages many crop hosts worldwide and is responsible for heavy economic losses. Early diagnosis and detection of the disease would allow for more effective crop management practices to prevent outbreaks in field or greenhouse settings. Furthermore, having a simple, non-invasive way to quantify the extent of gray mold disease is important for plant pathologists interested in measuring infection rates. In this paper, we design and build a bispectral imaging system for discriminating between leaf regions infected with gray mold and those that remain unharmed on a lettuce ( Lactuca spp.) host. First, we describe a method to select two optimal (high contrast) spectral bands from continuous hyperspectral imagery (450-800 nm). We then explain the process of building a system based on these two spectral bands, located at 540 and 670 nm. The resultant system uses two cameras, with a narrow band-pass spectral filter mounted on each, to measure the bispectral reflectance of a lettuce leaf. The two resulting images are combined using a normalized difference calculation that produces a single image with high contrast between the leaves' infected and healthy regions. A classifier was then created based on the thresholding of single pixel values. We demonstrate that this simple classification produces a true-positive rate of 95.25% with a false-positive rate of 9.316% in laboratory conditions.
Abstract Flower opening and closure are traits of reproductive importance in all angiosperms because they determine the success of self- and cross-pollination. The temporal nature of this phenotype rendered it a difficult target for genetic studies. Cultivated and wild lettuce, Lactuca spp., have composite inflorescences that open only once. An L. serriola×L. sativa F6 recombinant inbred line (RIL) population differed markedly for daily floral opening time. This population was used to map the genetic determinants of this trait; the floral opening time of 236 RILs was scored using time-course image series obtained by drone-based phenotyping on two occasions. Floral pixels were identified from the images using a support vector machine with an accuracy >99%. A Bayesian inference method was developed to extract the peak floral opening time for individual genotypes from the time-stamped image data. Two independent quantitative trait loci (QTLs; Daily Floral Opening 2.1 and qDFO8.1) explaining >30% of the phenotypic variation in floral opening time were discovered. Candidate genes with non-synonymous polymorphisms in coding sequences were identified within the QTLs. This study demonstrates the power of combining remote sensing, machine learning, Bayesian statistics, and genome-wide marker data for studying the genetics of recalcitrant phenotypes.
Research has been increasingly focusing on the selection of novel and effective biological control agents (BCAs) against soil-borne plant pathogens. The large-scale application of BCAs requires fast and robust screening methods for the evaluation of the efficacy of high numbers of candidates. In this context, the digital technologies can be applied not only for early disease detection but also for rapid performance analyses of BCAs. The present study investigates the ability of different Trichoderma spp. to contain the development of main baby-leaf vegetable pathogens and applies functional plant imaging to select the best performing antagonists against multiple pathosystems. Specifically, sixteen different Trichoderma spp. strains were characterized both in vivo and in vitro for their ability to contain R. solani, S. sclerotiorum and S. rolfsii development. All Trichoderma spp. showed, in vitro significant radial growth inhibition of the target phytopathogens. Furthermore, biocontrol trials were performed on wild rocket, green and red baby lettuces infected, respectively, with R. solani, S. sclerotiorum and S. rolfsii . The plant status was monitored by using hyperspectral imaging. Two strains, Tl35 and Ta56, belonging to T. longibrachiatum and T. atroviride species, significantly reduced disease incidence and severity (DI and DSI) in the three pathosystems. Vegetation indices, calculated on the hyperspectral data extracted from the images of plant- Trichoderma -pathogen interaction, proved to be suitable to refer about the plant health status. Four of them (OSAVI, SAVI, TSAVI and TVI) were found informative for all the pathosystems analyzed, resulting closely correlated to DSI according to significant changes in the spectral signatures among health, infected and bio-protected plants. Findings clearly indicate the possibility to promote sustainable disease management of crops by applying digital plant imaging as large-scale screening method of BCAs' effectiveness and precision biological control support.
Deep understanding of genetic architecture of water-stress tolerance is critical for efficient and optimal development of water-stress tolerant cultivars, which is the most economical and environmentally sound approach to maintain lettuce production with limited irrigation. Lettuce ( Lactuca sativa L.) production in areas with limited precipitation relies heavily on the use of ground water for irrigation. Lettuce plants are highly susceptible to water-stress, which also affects their nutrient uptake efficiency. Water stressed plants show reduced growth, lower biomass, and early bolting and flowering resulting in bitter flavors. Traditional phenotyping methods to evaluate water-stress are labor intensive, time-consuming and prone to errors. High throughput phenotyping platforms using kinetic chlorophyll fluorescence and hyperspectral imaging can effectively attain physiological traits related to photosynthesis and secondary metabolites that can enhance breeding efficiency for water-stress tolerance. Kinetic chlorophyll fluorescence and hyperspectral imaging along with traditional horticultural traits identified genomic loci affected by water-stress. Supervised machine learning models were evaluated for their accuracy to distinguish water-stressed plants and to identify the most important water-stress related parameters in lettuce. Random Forest (RF) had classification accuracy of 89.7% using kinetic chlorophyll fluorescence parameters and Neural Network (NN) had classification accuracy of 89.8% using hyperspectral imaging derived vegetation indices. The top ten chlorophyll fluorescence parameters and vegetation indices selected by sequential forward selection by RF and NN were genetically mapped using a L. sativa × L. serriola interspecific recombinant inbred line (RIL) population. A total of 25 quantitative trait loci (QTL) segregating for water-stress related horticultural traits, 26 QTL for the chlorophyll fluorescence traits and 34 QTL for spectral vegetation indices (VI) were identified. The percent phenotypic variation (PV) explained by the horticultural QTL ranged from 6.41 to 19.5%, PV explained by chlorophyll fluorescence QTL ranged from 6.93 to 13.26% while the PV explained by the VI QTL ranged from 7.2 to 17.19%. Eight QTL clusters harboring co-localized QTL for horticultural traits, chlorophyll fluorescence parameters and VI were identified on six lettuce chromosomes. Molecular markers linked to the mapped QTL clusters can be targeted for marker-assisted selection to develop water-stress tolerant lettuce.
Reproduction assets found保存済みの本文根拠を更新済みルールで再検証し、公開資産1件を確認しました。Supplement · publicSupplementary Table 1
Phenotypic values of selected chlorophyll fluorescence parameters and vegetation indices during drought stress progression.Open asset ↗lines:1465-1521Plant phenotyping relevance matchCrossref · bioRxiv · checked 13 Sept 2026
Yangminghao Liu · Daniel Patko · Ilonka Engelhardt · Timothy S George · Nicola Stanley-Wall · Vincent Ladmiral · Bruno Ameduri · Tim J Daniell · Nicola Holden · Michael P MacDonald · Lionel X Dupuy
Abstract Our understanding of plant-microbe interactions in soil is limited by the difficulty of observing processes at the microscopic scale throughout plants’ large volume of influence. Here, we present the development of 3D live microscopy for resolving plant-microbe interactions across the environment of an entire seedling growing in a transparent soil in tailor-made mesocosms, maintaining physical conditions for the culture of both plants and microorganisms. A tailor made dual-illumination light-sheet system acquired scattering signals from the plant whilst fluorescence signals were captured from transparent soil particles and labelled microorganisms, allowing the generation of quantitative data on samples approximately 3600 mm 3 in size with as good as 5 μm resolution at a rate of up to one scan every 30 minutes. The system tracked the movement of Bacillus subtilis populations in the rhizosphere of lettuce plants in real time, revealing previously unseen patterns of activity. Motile bacteria favoured small pore spaces over the surface of soil particles, colonising the root in a pulsatile manner. Migrations appeared to be directed towards the root cap, the point “first contact”, before subsequent colonisation of mature epidermis cells. Our findings show that microscopes dedicated to live environmental studies present an invaluable tool to understand plant-microbe interactions.
Plant phenotyping plays a crucial role in understanding variations in the phenotype of individual plants affected by environment, management, and genotype. Measurement of seed germination is an important phenotyping stage as germination impacts on the whole plant growth process. However, germination measurement has been limited to germination percentage of a seed population. Understanding of the germination time, from sowing to outbreak of the radicle from seed coat, at a single seed level is essential. How individual germination time and further plant growth are affected by its microenvironment and management factors remains elusive. Plant phenotype measurement system was developed to assess individual germination time of romaine lettuce (Lactuca sativa L. var. longifolia), using time-series two-dimensional camera images, and to analyze how microenvironment (volumetric water percent in seed tray, individual seed surface temperature and air temperature) and management factors (coated/uncoated seeds) affect the germination time for plant cohort research, emphasizing practicality in commercial cultivation. Germination experiments were conducted to demonstrate the performance of the system and its applicability for a whole plant growth process in a plant factory for commercial production and/or breeding. The developed phenotyping platform revealed the effects of microenvironment and management factors on germination time of individual seeds.
The yield and quality of fresh lettuce can be determined from the growth rate and color of individual plants. Manual assessment and phenotyping for hundreds of varieties of lettuce is very time consuming and labor intensive. In this study, we utilized a "Sensor-to-Plant" greenhouse phenotyping platform to periodically capture top-view images of lettuce, and datasets of over 2000 plants from 500 lettuce varieties were thus captured at eight time points during vegetative growth. Here, we present a novel object detection-semantic segmentation-phenotyping method based on convolutional neural networks (CNNs) to conduct non-invasive and high-throughput phenotyping of the growth and development status of multiple lettuce varieties. Multistage CNN models for object detection and semantic segmentation were integrated to bridge the gap between image capture and plant phenotyping. An object detection model was used to detect and identify each pot from the sequence of images with 99.82% accuracy, semantic segmentation model was utilized to segment and identify each lettuce plant with a 97.65% F1 score, and a phenotyping pipeline was utilized to extract a total of 15 static traits (related to geometry and color) of each lettuce plant. Furthermore, the dynamic traits (growth and accumulation rates) were calculated based on the changing curves of static traits at eight growth points. The correlation and descriptive ability of these static and dynamic traits were carefully evaluated for the interpretability of traits related to digital biomass and quality of lettuce, and the observed accumulation rates of static straits more accurately reflected the growth status of lettuce plants. Finally, we validated the application of image-based high-throughput phenotyping through geometric measurement and color grading for a wide range of lettuce varieties. The proposed method can be extended to crops such as maize, wheat, and soybean as a non-invasive means of phenotype evaluation and identification.
This work introduces a method that combines remote sensing and deep learning into a framework that is tailored for accurate, reliable and efficient counting and sizing of plants in aerial images. The investigated task focuses on two low-density crops, potato and lettuce. This double objective of counting and sizing is achieved through the detection and segmentation of individual plants by fine-tuning an existing deep learning architecture called Mask R-CNN. This paper includes a thorough discussion on the optimal parametrisation to adapt the Mask R-CNN architecture to this novel task. As we examine the correlation of the Mask R-CNN performance to the annotation volume and granularity (coarse or refined) of remotely sensed images of plants, we conclude that transfer learning can be effectively used to reduce the required amount of labelled data. Indeed, a previously trained Mask R-CNN on a low-density crop can improve performances after training on new crops. Once trained for a given crop, the Mask R-CNN solution is shown to outperform a manually-tuned computer vision algorithm. Model performances are assessed using intuitive metrics such as Mean Average Precision (mAP) from Intersection over Union (IoU) of the masks for individual plant segmentation and Multiple Object Tracking Accuracy (MOTA) for detection. The presented model reaches an mAP of 0.418 for potato plants and 0.660 for lettuces for the individual plant segmentation task. In detection, we obtain a MOTA of 0.781 for potato plants and 0.918 for lettuces.
Han, R. · Wong, A. J. Y. · Tang, Z. · Truco, M. J. · Lavelle, D. O. · Kozik, A. · Jin, Y. · Michelmore, R. W.
LettuceAerial / UAVFlowerSegmentationGrowth / time-series analysisGrowth / development / phenology
Flower opening and closure are traits of reproductive importance in all angiosperms because they determine the success of self- and cross-pollination. The temporal nature of this phenotype rendered it a difficult target for genetic studies. Cultivated and wild lettuce, Lactuca spp., have composite inflorescences comprised of multiple florets that open only once. Different accessions were observed to flower at different times of day. An F6 recombinant inbred line population (RIL) had been derived from accessions of L. serriola x L. sativa that originated from different environments and differed markedly for daily floral opening time. This population was used to map the genetic determinants of this trait; the floral opening time of 236 RILs was scored over a seven-hour period using time-course image series obtained by drone-based remote phenotyping on two occasions, one week apart. Floral pixels were identified from the images using a support vector machine (SVM) machine learning algorithm with an accuracy above 99%. A Bayesian inference method was developed to extract the peak floral opening time for individual genotypes from the time-stamped image data. Two independent QTLs, qDFO2.1 (Daily Floral Opening 2.1) and qDFO8.1, were discovered. Together, they explained more than 30% of the phenotypic variation in floral opening time. Candidate genes with non-synonymous polymorphisms in coding sequences were identified within the QTLs. This study demonstrates the power of combining remote imaging, machine learning, Bayesian statistics, and genome-wide marker data for studying the genetics of recalcitrant phenotypes such as floral opening time. One sentence summaryMachine learning and Bayesian analyses of drone-mediated remote phenotyping data revealed two genetic loci regulating differential daily flowering time in lettuce (Lactuca spp.).
Reproduction assets foundThe paper's Data Availability statement provides two paper-specific public assets: authors' analysis scripts (machine learning and Bayesian inference) on GitHub, and the GPS-anchored drone aerial image data on HydroShare. Both are directly used for this paper's phenotyping and analysis.Code · public51 2015-51181-24283 to RWM.
452
453 Data Availability
454 GBS data of the RILs and WGS data of the parents are available on the NCBI SRA database under
455 BioProjects PRJNA642889, PRJNA510128, and PRJNA478460, respectively. Scripts used in the
456 study for machine learning and Bayesian inference are available on GitHub at
457 https://www.github.com/rkbhan/FloralOpening. GPS-anchored aerial image data are available on
458 HydroShare at https://www.hydroshare.org/resource/1c5855dbeb3c49a8b5779300550e08f1/.
20Open asset ↗rkbhan/FloralOpeningpdf-layout-page:20 lines:1-57Dataset · publicavailable on the NCBI SRA database under
455 BioProjects PRJNA642889, PRJNA510128, and PRJNA478460, respectively. Scripts used in the
456 study for machine learning and Bayesian inference are available on GitHub at
457 https://www.github.com/rkbhan/FloralOpening. GPS-anchored aerial image data are available on
458 HydroShare at https://www.hydroshare.org/resource/1c5855dbeb3c49a8b5779300550e08f1/.
20Open asset ↗pdf-layout-page:20 lines:1-57Plant phenotyping relevance matchCrossref · checked 15 Sept 2026
Published1 Jul 2020IOP Conference Series: Earth and Environmental ScienceCited by 11 · OpenAlex ↗
Abstract The plant factory is extensive cultivation to produce a high quality of vegetables under a controllable environment. The concept of Precision Agriculture (PA) was introduced to improve the plant factory production by the implementation of a crop growth monitoring system. Crop growth can be estimated by monitoring of crop height and canopy foliage by the use of computer vision technology. In our previous study, we have introduced a plant height monitoring system based on depth perception using a stereo camera. However, the validity of the various type of leave is necessary to be tested. The objective of this study was to implement the crop growth monitoring system to monitor plant development with various type of leave for system validation and evaluation. The crop growth monitoring system composed of a stereo camera implementing the depth perception for estimating the distance from camera to highest point in the crop. The implementation of the system with various types of leaves and characteristics has been conducted for (a) Samhon, (b) Lettuce, and (c) Pagoda. The developed crop growth monitoring system could perform the time series estimation of crop height with a maximum error of RMSE 0.875 cm on Pagoda, and MAPE of 5.56% on Lettuce. The system demonstrates better estimation on Samhong with minimum error RMSE of 0.408cm and 2.27 % of MAPE. Overall validation for the estimated height vs actual measurement indicates that the coefficient of determination higher than 0.7 means that it has substantial features for estimating the plant height.
Proximal sensors in controlled environment agriculture (CEA) are used to monitor plant growth, yield, and water consumption with non-destructive technologies. Rapid and continuous monitoring of environmental and crop parameters may be used to develop mathematical models to predict crop response to microclimatic changes. Here, we applied the energy cascade model (MEC) on green- and red-leaf butterhead lettuce ( Lactuca sativa L. var. capitata ). We tooled up the model to describe the changing leaf functional efficiency during the growing period. We validated the model on an independent dataset with two different vapor pressure deficit (VPD) levels, corresponding to nominal (low VPD) and off-nominal (high VPD) conditions. Under low VPD, the modified model accurately predicted the transpiration rate (RMSE = 0.10 Lm -2 ), edible biomass (RMSE = 6.87 g m -2 ), net-photosynthesis (rBIAS = 34%), and stomatal conductance (rBIAS = 39%). Under high VPD, the model overestimated photosynthesis and stomatal conductance (rBIAS = 76-68%). This inconsistency is likely due to the empirical nature of the original model, which was designed for nominal conditions. Here, applications of the modified model are discussed, and possible improvements are suggested based on plant morpho-physiological changes occurring in sub-optimal scenarios.
Gray mold disease caused by the fungus Botrytis cinerea damages many crop hosts worldwide and is responsible for heavy economic losses. Early diagnosis and detection of the disease would allow for more effective crop management practices to prevent outbreaks in field or greenhouse settings. Furthermore, having a simple, non-invasive way to quantify the extent of gray mold disease is important for plant pathologists interested in quantifying infection rates. In this paper, we design and build a multispectral imaging system for discriminating between leaf regions, infected with gray mold, and those that remain unharmed on a lettuce ( Lactuca spp.) host. First, we describe a method to select two optimal (high contrast) spectral bands from continuous hyperspectral imagery (450-800 nm). We then built a system based on these two spectral bands, located at 540 and 670 nm. The resultant system uses two cameras, with a narrow band-pass spectral filter mounted on each, to measure the multispectral reflectance of a lettuce leaf. The two resulting images are combined using a normalized difference calculation that produces a single image with high contrast between the leaves’ infected and healthy regions. A classifier was then created based on the thresholding of single pixel values. We demonstrate that this simple classification produces a true positive rate of 95.25% with a false positive rate of 9.316%.
According to the Commission Regulation (EC) No. 1258/2011, the maximum allowed nitrate content of lettuce is defined within a broad range (2000-5000 mg NO 3 /kg), depending on harvest season and technology. This study focuses on the identification of the differences in nitrate accumulation between lettuce types and varieties, depending on production technology and on the investigation of the application of non-destructive FT-NIR spectroscopy for nitrate quantification, towards widely used UV-Vis spectroscopy. In the present study, combinations of seasons and technologies (spring × greenhouse, autumn × open field) were employed for the production of types (batavia, butterhead, lollo and oak leaf; both red and green colored); a total of 266 lettuce heads were analyzed. It was found that with standardized technology and conditions, autumn harvested green oak leaf lettuce types accumulated significantly less nitrate, than red oak or lollo leaf types. With spring harvested lettuces, batavia types generally accumulated generally more nitrates than butterhead types. Based on the linear discriminant analysis (LDA) of FT-NIR measurements the four distinct variety types diverge; the lollo type explicitly diverges from batavia and butterhead types. The LDA further revealed, that within lollo and oak leaf variety types, red and green leaved varieties diverge as well. A model was successfully built for the FT-NIR quantification of the nitrate content of lettuce samples (R 2 = 0.95; RMSEE = 74.4 mg/kg fresh weight; Q 2 = 0.90; RMSECV = 99.4 mg/kg fresh weight). The developed model is capable of the execution of a fast and non-invasive measurement; the method is suitable for the routine measurement of nitrate content in lettuce.
In order to train the neural network for plant phenotyping, a sufficient amount of training data must be prepared, which requires time-consuming manual data annotation process that often becomes the limiting step. Here, we show that an instance segmentation neural network aimed to phenotype the barley seed morphology of various cultivars, can be sufficiently trained purely by a synthetically generated dataset. Our attempt is based on the concept of domain randomization, where a large amount of image is generated by randomly orienting the seed object to a virtual canvas. The trained model showed 96% recall and 95% average Precision against the real-world test dataset. We show that our approach is effective also for various crops including rice, lettuce, oat, and wheat. Constructing and utilizing such synthetic data can be a powerful method to alleviate human labor costs for deploying deep learning-based analysis in the agricultural domain.
Reproduction assets foundThe authors publicly release both the synthetic and real-world seed image datasets and the analysis code (Mask R-CNN deployment and multivariate analysis notebooks) via their GitHub repository, explicitly stated in Data availability and Code availability sections.Dataset · publicSynthetically generated and real-world datasets can be obtained from the following GitHub repository ( https://github.com/totti0223/crop_seed_instance_segmentation ).Open asset ↗https://github.com/totti0223/crop_seed_instance_segmentationlines:149-171Code · publicCode to reproduce the deployment of the trained Mask R-CNN and multivariate analysis is formatted as IPython notebooks and can also be obtained from the GitHub repository ( https://github.com/totti0223/crop_seed_instance_segmentation ).Open asset ↗https://github.com/totti0223/crop_seed_instance_segmentationlines:149-171Plant phenotyping relevance matchEurope PMC · checked 9 Sept 2026
Abstract Background : Vegetables are one of the most important nitrate sources of human diary diet. Establishing fast and accurate in situ nitrate monitoring approaches that could be used in the plant growth process and vegetable markets is essential. Results: Incorporating the unique feature of N-O asymmetric stretch absorption in the mid-infrared region (1500-1200 cm -1 ), portable Fourier-transform infrared attenuated total reflectance (FTIR-ATR) spectroscopic instruments, along with the Euclidean distance-modified intelligent algorithm extreme learning machine (ED-ELM) model, were employed to evaluate the nitrate contents in leafy vegetables. A total of 1224 samples of four popular vegetables (Chinese cabbage, swamp cabbage, celery, and lettuce) were analyzed. The results indicated that the nitrate contents (mean values: Chinese cabbage: 7550 mg/kg; swamp cabbage: 4219 mg/kg; celery: 4164 mg/kg; lettuce: 4322 mg/kg) highly exceeded the World Health Organization (WHO))-specified maximum tolerance limits. The ED-ELM model showed a better performance with the root-mean-square-error of 799.7 mg/kg, the determination coefficients of 0.93, the ratio of performance to deviation of 2.22, the optimized calibration dataset number of 100, and the number of hidden neurons of 30. Conclusion: The results confirmed that FTIR-ATR, along with the suitable model algorithms, could be used as a potential rapid and accurate method to monitor the nitrate contents in the fields of agriculture and food safety.
Lettuce (Lactuca sativa L.) is a popular leafy vegetable, widely grown and consumed throughout the world. Growing Lettuce plants in controlled environment, it is useful to increase the yield and obtain production year-round. In CEA (Controlled Environment Agriculture), computer technology is an integral part in the production and different sensors used to monitor environmental parameters and activate environmental control, are necessary. With the advent of technology, proximal sensors and plant phenotyping (in terms of physiological measurements of plant status) can help farmers in crop management. However, these kinds of tools are often expensive or inaccessible for stakeholders. The application of these tools to small-scale cultivation trials, could provide data for the implementation of mathematical models capable of predicting changes possibly happening during the cultivation. These models could then be applied at larger scales, as extensive farm production and be used to help in the cultivation management. In this study, green and red cultivars of Lactuca sativa L. ‘Salanova’ were grown in a growth chamber under controlled environmental condition (T, RH, light intensity and quality) in two trials under different vapour pressure deficit (VPD) : 1) VPD of 0.70 kPa (Low VPD; nominal condition) and 2) VPD of 1.76 (High VPD; off nominal condition). Plants were irrigated to field-capacity and weighted every-day in order to record daily ET; infra-red measurements were carried out to record leaf temperature and pictures were taken to monitor growth during the cultivation. Furthermore, after 23 days, on fully developed leaves, eco-physiological analyses (gas exchange and chlorophyll “a” measurements) were performed to assess the plant physiological behaviour in response to the different environmental conditions. Environmental data, were used as inputs in an energy cascade model (MEC) to predict changes in the plant daily growth, photosynthesis and evapotranspiration. The original model, was implemented with a few variations: leaf temperature (T) was used in place of air T for computing the stomatal conductance (gs) and the model parameters maxCUE and maxQY, were differentiated for the nominal and off-nominal scenarios and for green and red lettuce cultivars. After the validation against experimental data, this model appears to be a promising tool that can be implemented for forecasting variations triggered by anomalies in the environmental control. However, a next step will be to add a few parameters that will consider the intrinsic morpho-physiological variability of plants during leaf development.
Hyperspectral imaging for agricultural applications provides a solution for non-destructive, large-area crop monitoring. However, current products are bulky and expensive due to complicated optics and electronics. A linear variable filter was developed for implementation into a prototype hyperspectral imaging camera that demonstrates good spectral performance between 450 and 900 nm. Equipped with a feature extraction and classification algorithm, the proposed system can be used to determine potato plant health with ∼88 % accuracy. This algorithm was also capable of species identification and is demonstrated as being capable of differentiating between rocket, lettuce, and spinach. Results are promising for an entry-level, low-cost hyperspectral imaging solution for agriculture applications.
High plant number per unit area makes it challenging to monitor plant growth in controlled environment agriculture (CEA) systems. Our objective was to develop and validate image analysis technique that uses a smartphone connected to local desktop computer for non-destructive measurement of growth characteristics of several species commonly grown in CEA. Using mobile apps, an iPhone-6 was remotely connected to a local computer containing image-processing software (MATLAB) and script. Smartphone was used to capture images of plants belonging to several species including basil, leaf lettuce, tomato, and zinnia. The images were moved to a folder on cloud storage and remotely processed on a local computer to derive estimated leaf area (LAₑₛₜᵢₘₐₜₑd) of plants. Regression analysis indicated a near perfect linear relation between measured leaf area (LAₘₑₐₛᵤᵣₑd) and LAₑₛₜᵢₘₐₜₑd (r² = 0.98) and shoot dry weight (SDW) and LAₑₛₜᵢₘₐₜₑd (r² = 0.94) when data were pooled from all species. No significant differences were observed when relative growth rate (RGR) was measured using either SDW or LAₑₛₜᵢₘₐₜₑd values. Further, results indicated that real-time and non-invasive LAₑₛₜᵢₘₐₜₑd measurements can be used to track plant growth differences over time. This method was able to identify plant growth differences more accurately than visual assessments on plants. Our findings indicate that LAₑₛₜᵢₘₐₜₑd can be used for accurate and non-invasive measurement of growth characteristics of plants in academic research. The technique can also aid in maximizing productivity, minimizing resource wastage and harvesting crops timely in commercial production.
Incorporating deep learning in the image analysis pipeline has opened the possibility of introducing precision phenotyping in the field of agriculture. However, to train the neural network, a sufficient amount of training data must be prepared, which requires a time-consuming manual data annotation process that often becomes the limiting step. Here, we show that an instance segmentation neural network (Mask R-CNN) aimed to phenotype the barley seed morphology of various cultivars, can be sufficiently trained purely by a synthetically generated dataset. Our attempt is based on the concept of domain randomization , where a large amount of image is generated by randomly orienting the seed object to a virtual canvas. After training with such a dataset, performance based on recall and the average Precision of the real-world test dataset achieved 96% and 95%, respectively. Applying our pipeline enables extraction of morphological parameters at a large scale, enabling precise characterization of the natural variation of barley from a multivariate perspective. Importantly, we show that our approach is effective not only for barley seeds but also for various crops including rice, lettuce, oat, and wheat, and thus supporting the fact that the performance benefits of this technique is generic. We propose that constructing and utilizing such synthetic data can be a powerful method to alleviate human labor costs needed to prepare the training dataset for deep learning in the agricultural domain.
Societal Impact Statement Advancements in our ability to rapidly detect plant responses to stress are necessary to improve crop management practices and meet the global challenge of food security. Using optical approaches to detect plant stress before symptoms become apparent has great potential, but these approaches lack testing in multiple‐stress environments and fail to fully exploit the data collected. Using hyperspectral data from lettuce, we show that optical measurements can provide growers with important stress‐related information to inform crop management practices. We suggest that integrating this technology into protected agrosystems, such as greenhouses, could greatly improve crop quality and yield. Summary Tools to detect and predict stress pre‐visually are essential to optimally manage agrosystems. Here, we investigated the capability of reflectance spectroscopy to characterize responses of asymptomatic crop leaves under multi‐stress conditions. Full range (350–2,500 nm) reflectance measurements and traditional plant stress responses were collected on lettuce leaves under the combination of different supplemental light types and intensities, fertilization and salinity. Partial least‐squares discriminate analysis and regression modeling and spectral indices were employed to characterize plant responses to multiple stress conditions, both alone and in combination. Spectral profiles (400–800 nm + 1,900–2,200 nm) of individuals grown under variable environments were statistically different ( p R 2 : 0.70–0.84). Higher lettuce yield and quality was found under sodium light at high intensity (850 µmol m −2 s −1 photosynthetic active radiation), with high fertilization (150 ppm N) and no salinity. Our findings highlight the utility and limitations of vegetation spectroscopy in a protected agrosystem. We suggest that integration of vegetation spectroscopy into intelligent and automated greenhouses and other protected systems could enhance management efficiency, as well as crop quality and yield.
Aerial imagery is regularly used by crop researchers, growers and farmers to monitor crops during the growing season. To extract meaningful information from large-scale aerial images collected from the field, high-throughput phenotypic analysis solutions are required, which not only produce high-quality measures of key crop traits, but also support professionals to make prompt and reliable crop management decisions. Here, we report AirSurf, an automated and open-source analytic platform that combines modern computer vision, up-to-date machine learning, and modular software engineering in order to measure yield-related phenotypes from ultra-large aerial imagery. To quantify millions of in-field lettuces acquired by fixed-wing light aircrafts equipped with normalised difference vegetation index (NDVI) sensors, we customised AirSurf by combining computer vision algorithms and a deep-learning classifier trained with over 100,000 labelled lettuce signals. The tailored platform, AirSurf- Lettuce , is capable of scoring and categorising iceberg lettuces with high accuracy (>98%). Furthermore, novel analysis functions have been developed to map lettuce size distribution across the field, based on which associated global positioning system (GPS) tagged harvest regions have been identified to enable growers and farmers to conduct precision agricultural practises in order to improve the actual yield as well as crop marketability before the harvest.
Productivity stabilization is a critical issue facing plant factories. As such, researchers have been investigating growth prediction with the overall goal of improving productivity. The projected area of a plant (PA) is usually used for growth prediction, by which the growth of a plant is estimated by observing the overall approximate movement of the plant. To overcome this problem, this study focused on the time-series movement of plant leaves, using optical flow (OF) analysis to acquire this information for a lettuce. OF analysis is an image processing method that extracts the difference between two consecutive frames caused by the movement of the subject. Experiments were carried out at a commercial large-scale plant factory. By using a microcomputer with a camera module placed above the lettuce seedlings, images of 338 seedlings were taken every 20 min over 9 days (from the 6th to the 15th day after sowing). Then, the features of the leaf movement were extracted from the image by calculating the normal-vector in the OF analysis, and these features were applied to machine learning to predict the fresh weight of the lettuce at harvest time (38 days after sowing). The growth prediction model using the features extracted from the OF analysis was found to perform well with a correlation ratio of 0.743. Furthermore, this study also considered a phenotyping system that was capable of automatically analyzing a plant image, which would allow this growth prediction model to be widely used in commercial plant factories.
Abstract Aerial imagery is regularly used by farmers and growers to monitor crops during the growing season. To extract meaningful phenotypic information from large-scale aerial images collected regularly from the field, high-throughput analytic solutions are required, which not only produce high-quality measures of key crop traits, but also support agricultural practitioners to make reliable management decisions of their crops. Here, we report AirSurf- Lettuce , an automated and open-source aerial image analysis platform that combines modern computer vision, up-to-date machine learning, and modular software engineering to measure yield-related phenotypes of millions of lettuces across the field. Utilising ultra-large normalized difference vegetation index (NDVI) images acquired by fixed-wing light aircrafts together with a deep-learning classifier trained with over 100,000 labelled lettuce signals, the platform is capable of scoring and categorising iceberg lettuces with high accuracy (>98%). Furthermore, novel analysis functions have been developed to map lettuce size distribution in the field, based on which global positioning system (GPS) tagged harvest regions can be derived to enable growers and farmers’ precise harvest strategies and marketability estimates before the harvest.
Shen T, Kong W, Liu F, Chen Z, Yao J, Wang W, Peng J, Chen H, He Y.
LettuceRaman / spectroscopyLeafClassification
Quick access to cadmium (Cd) contamination in lettuce is important to supervise the leafy vegetable growth environment and market. This study aims to apply laser-induced breakdown spectroscopy (LIBS) technology for fast determination of Cd content and diagnosis of the Cd contamination degree in lettuce. Emission lines Cd II 214.44 nm, Cd II 226.50 nm, and Cd I 228.80 nm were selected to establish the univariate analysis model. Multivariate analysis including partial least squares (PLS) regression, was used to establish Cd content calibration models, and PLS model based on 22 variables selected by genetic algorithm (GA) obtained the best performance with correlation coefficient in the prediction set Rp ² = 0.9716, limit of detection ( LOD ) = 1.7 mg/kg. K-Nearest Neighbors (KNN) and random forest (RF) were used to analyze Cd contamination degree, and RF model obtained the correct classification rate of 100% in prediction set. The preliminary results indicate LIBS coupled with chemometrics could be used as a fast, efficient and low-cost method to assess Cd contamination in the vegetable industry.
Monitoring the health and yield of crops during production is an important, but labour intensive component of commercial agriculture, especially in high value crop such as lettuce. This article proposes a novel method for segmenting lettuce in coloured 3D point clouds and estimating the fresh weight. The proposed segmentation method operates by clustering points into leaves and then evaluating their affiliation to a lettuce of interest. From the segmented lettuce point clouds, the volume, surface area, leaf cover area and height predictors are extracted and correlated to the fresh weight. The proposed segmentation and yield estimation methods are evaluated on Cos and Iceberg lettuce point clouds generated from images collected by an agricultural robot in an outdoor field experiment. The results demonstrate that the proposed segmentation method is able to successfully isolate lettuce (F1-score = 0.88–0.91). Analysis of the segmented lettuce models show that the calculated surface areas correlate strongly with measured fresh weight (R2 = 0.84–0.94). Not only does this validate the segmentation method, it allows an accurate estimate of the lettuce fresh weight (RMSE = 27–50 g) to be produced non-destructively.
The crop water stress index (CWSI) has been shown to be a tool that could be used for non-contact and real-time monitoring of plant water status, which is a key requirement for the precision irrigation management of crops. However, its adoption for irrigation scheduling is limited because of the need to know the baseline temperatures which are required for its calculation. In this study, the canopy temperature of greenhouse cultivated lettuce plants which were maintained as either well-watered or non-transpiring was continuously monitored along with prevailing environmental conditions during a five week period. This data was applied in developing a dynamic model that can be used for predicting the baseline temperatures. Input variables for the dynamic model included air temperature, shortwave irradiance, and air vapour pressure deficit measured at a 10 s interval. During a follow up study, the dynamic model successfully predicted the baseline temperatures producing mean absolute errors (MAE) that varied between 0.17 °C and 0.29 °C, and root mean squared errors (RMSE) that varied between 0.21 °C and 0.35 °C when comparing model predictions with measured values. The model predicted baseline temperatures were applied in calculating an empirical CWSI for lettuce plants receiving one of two irrigation treatments. The empirical CWSI consistently differentiated between the irrigation treatments and was significantly correlated with the theoretical CWSI with correlation coefficient (r) values greater than 0.9. The dynamic model presented in this study requires easily measured input parameters for the prediction of the baseline temperatures. This eliminates the need to maintain artificial reference surfaces required in other empirical approaches for the CWSI calculation and also eliminates the need for computing the complex theoretical CWSI.
Non-destructive plant growth measurement is essential for plant growth and health research. As a 3D sensor, Kinect v2 has huge potentials in agriculture applications, benefited from its low price and strong robustness. The paper proposes a Kinect-based automatic system for non-destructive growth measurement of leafy vegetables. The system used a turntable to acquire multi-view point clouds of the measured plant. Then a series of suitable algorithms were applied to obtain a fine 3D reconstruction for the plant, while measuring the key growth parameters including relative/absolute height, total/projected leaf area and volume. In experiment, 63 pots of lettuce in different growth stages were measured. The result shows that the Kinect-measured height and projected area have fine linear relationship with reference measurements. While the measured total area and volume both follow power law distributions with reference data. All these data have shown good fitting goodness ( R ² = 0.9457-0.9914). In the study of biomass correlations, the Kinect-measured volume was found to have a good power law relationship ( R ² = 0.9281) with fresh weight. In addition, the system practicality was validated by performance and robustness analysis.
Reproduction assets foundThe paper's Supplementary Materials, available at the MDPI s1 URL, explicitly contain the paper-specific phenotyping assets: point clouds and meshes shown in figures, the datasets used for the scatter plots of Kinect-measured growth parameters vs. reference measurements, and interactive MATLAB 3D scatter plots. No codeDataset · publicThe following are available online at http://www.mdpi.com/1424-8220/18/3/806/s1 . Supplementary data associated with this article have been provided. These data include the point clouds and meshes appeared in figures, the data sets used by scatter plots, and interactive MATLAB 3D scatter plots.Open asset ↗lines:114-135Plant phenotyping relevance matchCrossref · checked 10 Sept 2026
This work introduces a new vision-based approach for estimating chlorophyll contents in a plant leaf using reflectance and transmittance as base parameters. Images of the top and underside of the leaf are captured. To estimate the base parameters (reflectance/transmittance), a novel optical arrangement is proposed. The chlorophyll content is then estimated by using linear regression where the inputs are the reflectance and transmittance of the leaf. Performance of the proposed method for chlorophyll content estimation was compared with a spectrophotometer and a Soil Plant Analysis Development (SPAD) meter. Chlorophyll content estimation was realized for Lactuca sativa L., Azadirachta indica, Canavalia ensiforme, and Lycopersicon esculentum. Experimental results showed that—in terms of accuracy and processing speed—the proposed algorithm outperformed many of the previous vision-based approach methods that have used SPAD as a reference device. On the other hand, the accuracy reached is 91% for crops such as Azadirachta indica, where the chlorophyll value was obtained using the spectrophotometer. Additionally, it was possible to achieve an estimation of the chlorophyll content in the leaf every 200 ms with a low-cost camera and a simple optical arrangement. This non-destructive method increased accuracy in the chlorophyll content estimation by using an optical arrangement that yielded both the reflectance and transmittance information, while the required hardware is cheap.
Background Plant science uses increasing amounts of phenotypic data to unravel the complex interactions between biological systems and their variable environments. Originally, phenotyping approaches were limited by manual, often destructive operations, causing large errors. Plant imaging emerged as a viable alternative allowing non-invasive and automated data acquisition. Several procedures based on image analysis were developed to monitor leaf growth as a major phenotyping target. However, in most proposals, a time-consuming parameterization of the analysis pipeline is required to handle variable conditions between images, particularly in the field due to unstable light and interferences with soil surface or weeds. To cope with these difficulties, we developed a low-cost, 2D imaging method, hereafter called PYM. The method is based on plant leaf ability to absorb blue light while reflecting infrared wavelengths. PYM consists of a Raspberry Pi computer equipped with an infrared camera and a blue filter and is associated with scripts that compute projected leaf area. This new method was tested on diverse species placed in contrasting conditions. Application to field conditions was evaluated on lettuces grown under photovoltaic panels. The objective was to look for possible acclimation of leaf expansion under photovoltaic panels to optimise the use of solar radiation per unit soil area. Results The new PYM device proved to be efficient and accurate for screening leaf area of various species in wide ranges of environments. In the most challenging conditions that we tested, error on plant leaf area was reduced to 5% using PYM compared to 100% when using a recently published method. A high-throughput phenotyping cart, holding 6 chained PYM devices, was designed to capture up to 2000 pictures of field-grown lettuce plants in less than 2 h. Automated analysis of image stacks of individual plants over their growth cycles revealed unexpected differences in leaf expansion rate between lettuces rows depending on their position below or between the photovoltaic panels. Conclusions The imaging device described here has several benefits, such as affordability, low cost, reliability and flexibility for online analysis and storage. It should be easily appropriated and customized to meet the needs of various users.
Reproduction assets foundThe authors explicitly state that the original pictures and PYM analysis code are publicly available in a GitHub repository. The lettuce agrivoltaic phenotype datasets are not public (confidentiality agreement; available on request), so only the code qualifies as a public paper-specific asset.Code · publicThe original pictures and code for PYM method used in the current study are available in the repository https://github.com/bevalle/pym .Open asset ↗bevalle/pymlines:174-250Plant phenotyping relevance matchEurope PMC · checked 15 Sept 2026
Davis RA, Rippner DA, Hausner SH, Parikh SJ, McElrone AJ, Sutcliffe JL.
LettuceMicroscopyMRI / PETLeafRootTracking
Engineered nanoparticles (NPs) are increasingly used in commercial products including automotive lubricants, clothing, deodorants, sunscreens, and cosmetics and can potentially accumulate in our food supply. Given their size it is difficult to detect and visualize the presence of NPs in environmental samples, including crop plants. New analytical tools are needed to fill the void for detection and visualization of NPs in complex biological and environmental matrices. We aimed to determine whether radiolabeled NPs could be used as a noninvasive, highly sensitive analytical tool to quantitatively track and visualize NP transport and accumulation in vivo in lettuce (Lactuca sativa) and to investigate the effect of NP size on transport and distribution over time using a combination of autoradiography, positron emission tomography (PET)/computed tomography (CT), scanning electron microscopy (SEM), and transition electron microscopy (TEM). Azide functionalized NPs were radiolabeled via a "click" reaction with copper-64 ( 64 Cu)-1,4,7-triazacyclononane triacetic acid (NOTA) azadibenzocyclooctyne (ADIBO) conjugate ([ 64 Cu]-ADIBO-NOTA) via copper-free Huisgen-1,3-dipolar cycloaddition reaction. This yielded radiolabeled [ 64 Cu]-NPs of uniform shape and size with a high radiochemical purity (>99%), specific activity of 2.2 mCi/mg of NP, and high stability (i.e., no detectable dissolution) over 24 h across a pH range of 5-9. Both PET/CT and autoradiography showed that [ 64 Cu]-NPs entered the lettuce seedling roots and were rapidly transported to the cotyledons with the majority of the accumulation inside the roots. Uptake and transport of intact NPs was size-dependent, and in combination with the accumulation within the roots suggests a filtering effect of the plant cell walls at various points along the water transport pathway.
Main conclusion Environmentally induced variation and the genotypic differences in flavonoid and phenolic content in lettuce can be reliably detected using the appropriate parameters derived from the records of rapid non-invasive fluorescence technique. The chlorophyll fluorescence excitation ratio method was designed as a rapid and non-invasive tool to estimate the content of UV-absorbing phenolic compounds in plants. Using this technique, we have assessed the dynamics of accumulation of flavonoids related to developmental changes and environmental effects. Moreover, we have tested appropriateness of the method to identify the genotypic differences and fluctuations in total phenolics and flavonoid content in lettuce. Six green and two red genotypes of lettuce (Lactuca sativa L.) grown in pots were exposed to two different environments for 50 days: direct sunlight (UV-exposed) and greenhouse conditions (low UV). The indices based on the measurements of chlorophyll fluorescence after red, green and UV excitation indicated increase of the content of UV-absorbing compounds and anthocyanins in the epidermis of lettuce leaves. In similar, the biochemical analyses performed at the end of the experiment confirmed significantly higher total phenolic and flavonoid content in lettuce plants exposed to direct sun compared to greenhouse conditions and in red compared to green genotypes. As the correlation between the standard fluorescence indices and the biochemical records was negatively influenced by the presence of red genotypes, we proposed the use of a new parameter named Modified Flavonoid Index (MFI) taking into an account both absorbance changes due to flavonol and anthocyanin content, for which the correlation with flavonoid and phenolic content was relatively good. Thus, our results confirmed that the fluorescence excitation ratio method is useful for identifying the major differences in phenolic and flavonoid content in lettuce plants and it can be used for high-throughput pre-screening and phenotyping of leafy vegetables in research and breeding applications towards improvement of vegetable health effects.
Rapid development of plants is important for the production of 'baby-leaf' lettuce that is harvested when plants reach the four- to eight-leaf stage of growth. However, environmental factors, such as high or low temperature, or elevated concentrations of salt, inhibit lettuce growth. Therefore, non-destructive evaluations of plants can provide valuable information to breeders and growers. The objective of the present study was to test the feasibility of using non-destructive phenotyping with optical sensors for the evaluations of lettuce plants in early stages of development. We performed the series of experiments to determine if hyperspectral imaging and chlorophyll fluorescence imaging can determine phenotypic changes manifested on lettuce plants subjected to the extreme temperature and salinity stress treatments. Our results indicate that top view optical sensors alone can accurately determine plant size to approximately 7 g fresh weight. Hyperspectral imaging analysis was able to detect changes in the total chlorophyll (RCC) and anthocyanin (RAC) content, while chlorophyll fluorescence imaging revealed photoinhibition and reduction of plant growth caused by the extreme growing temperatures (3 and 39°C) and salinity (100 mM NaCl). Though no significant correlation was found between F v /F m and decrease in plant growth due to stress when comparisons were made across multiple accessions, our results indicate that lettuce plants have a high adaptability to both low (3°C) and high (39°C) temperatures, with no permanent damage to photosynthetic apparatus and fast recovery of plants after moving them to the optimal (21°C) temperature. We have also detected a strong relationship between visual rating of the green- and red-leaf color intensity and RCC and RAC, respectively. Differences in RAC among accessions suggest that the selection for intense red color may be easier to perform at somewhat lower than the optimal temperature. This study serves as a proof of concept that optical sensors can be successfully used as tools for breeders when evaluating young lettuce plants. Moreover, we were able to identify the locus for light green leaf color ( qLG4 ), and position this locus on the molecular linkage map of lettuce, which shows that these techniques have sufficient resolution to be used in a genetic context in lettuce.
Poorly grown plants that result from differences in individuals lead to large profit losses for plant factories that use large electric power sources for cultivation. Thus, identifying and culling the low-grade plants at an early stage, using so-called seedlings diagnosis technology, plays an important role in avoiding large losses in plant factories. In this study, we developed a high-throughput diagnosis system using the measurement of chlorophyll fluorescence (CF) in a commercial large-scale plant factory, which produces about 5000 lettuce plants every day. At an early stage (6 days after sowing), a CF image of 7200 seedlings was captured every 4 h on the final greening day by a high-sensitivity CCD camera and an automatic transferring machine, and biological indices were extracted. Using machine learning, plant growth can be predicted with a high degree of accuracy based on biological indices including leaf size, amount of CF, and circadian rhythms in CF. Growth prediction was improved by addition of temporal information on CF. The present data also provide new insights into the relationships between growth and temporal information regulated by the inherent biological clock.
Robustness in lettuce, defined as the ability to produce stable yields across a wide range of environments, may be associated with below-ground traits such as water and nitrate capture. In lettuce, research on the role of root traits in resource acquisition has been rather limited. Exploring genetic variation for such traits and shoot performance in lettuce across environments can contribute to breeding for robustness. A population of 142 lettuce cultivars was evaluated during two seasons (spring and summer) in two different locations under organic cropping conditions, and water and nitrate capture below-ground and accumulation in the shoots were assessed at two sampling dates. Resource capture in each soil layer was measured using a volumetric method based on fresh and dry weight difference in the soil for soil moisture, and using an ion-specific electrode for nitrate. We used these results to carry out an association mapping study based on 1170 single nucleotide polymorphism markers. We demonstrated that our indirect, high-throughput phenotyping methodology was reliable and capable of quantifying genetic variation in resource capture. QTLs for below-ground traits were not detected at early sampling. Significant marker-trait associations were detected across trials for below-ground and shoot traits, in number and position varying with trial, highlighting the importance of the growing environment on the expression of the traits measured. The difficulty of identifying general patterns in the expression of the QTLs for below-ground traits across different environments calls for a more in-depth analysis of the physiological mechanisms at root level allowing sustained shoot growth.
Reproduction assets foundThe paper's phenotypic measurements (soil water/nitrate capture, shoot traits) and genotype scores/linkage map are stated to be available as Supplementary Material (Tables S1 and S2) hosted at the article's public Frontiers URL. No author analysis code or trained models are mentioned.Supplement · publicey also thank Jan Velema, Marcel van Diemen and Pieter Schwegman, Vitalis Organic Seeds, for providing seeds, advice, and insight. The project was financially supported through the Top Institute Green Genetics (project number: 2CFD024RP).
Supplementary material
The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fpls.2016.00343
Click here for additional data file.
Click here for additional data file.
Click here for additional data file.
References
Biddington N. L., Dearman A. S. (1985). The effects of mechanically-induced stress on water loss and drought resistance in lettuce, cauliflower and celery seedlings. Ann. Bot.
56, 795–8Open asset ↗lines:840-871Plant phenotyping relevance matchEurope PMC · checked 15 Sept 2026
Published1 Feb 2016Guang pu xue yu guang pu fen xi = Guang pu
Nitrogen fertilizer is necessary to improve yield and quality of lettuce. Spectroscopy is one of the most effective techniques used to detect crop nitrogen content. In this study, canopy reflectance spectra were acquired under five levels of nitrogen, and then were Savitzky-Golay smoothed, the first-order derivative spectra were calculated from the smoothed spectra to eliminate noise effects. Backward interval partial least squares (BiPLS), genetic algorithm (GA) and successive projections algorithm (SPA) were combined to select the efficient wavelengths. The number of variables was decreased from 2,151 to 8. The optimal intervals or variables were used to build multivariable linear regression (MLR) model, radial basis function neural network (RBFNN) models and extreme learning machine (ELM) models. This work proved that the results of BiPLS-GA-SPA-ELM model was superior to others with RMSEC was 0.241 6%, Rc was 0.934 6, RMSEP was 0.284 2% and Rp was 0.921 8. Our research results may provide a foundation for nutrition regulation and developing instrument.