Plant 3D reconstruction based on LiDAR and multi-view sequence images

Abstract

The 3D reconstruction of plant based on LiDAR is the main way to obtain the spatial structure of plant rapidly, nondestructive and all-weather. However, the influence of LiDAR instrument performance and field operation environment, the point cloud data obtained will lose the details of plant and reduce the accuracy of the model. In this study, the three-dimensional point cloud of plants generated based on multi-view sequence images was taken as a reference. The optimized Iterative Closest Point registration was adopted to calibrate the point cloud data from the LiDAR scanning to improve the detailed characteristics of the plants and establish a 3D model of plant. At the same time, according to the measured plant phenotype parameters (leaf length, leaf width, leaf area, plant height), the accuracy of 3D model was evaluated. The results showed that high accuracy of 3D reconstruction was obtained based on LiDAR and multi-view image sequence method. There was a good agreement between measured and calculated leaf area, leaf length, leaf width and plant height with R 2 >0.8 for leaf area, RMSE 0.85 for leaf length, R 2 >0.95 for leaf width. There was no significant difference for each phenotypic parameter between measured and calculated data (ANOVA, P <0.05). This method provides a technical reference for the research and application of LiDAR in fine modeling of field crops. Keywords: LiDAR, multi-view sequence images, plant 3D reconstruction, accuracy evaluatione DOI: 10.33440/j.ijpaa.20180101.0007 Citation: Wu J W, Xue X Y, Zhang S C, Qin W C, Chen C, Sun T. Plant 3D reconstruction based on LiDAR and multi-view sequence images. Int J Precis Agric Aviat, 2018; 1(1): 37–43.

Code and data availability

The article describes LiDAR/multi-view 3D reconstruction of plants with phenotype accuracy evaluation, but contains no data availability statement, no public dataset or image deposit, and no author code, scripts, models, or supplement with a public URL. Only commercial software (Agisoft) and cited prior work are named.

No evidence-backed public reproduction asset is currently recorded.