LITERAL, a portable, lightweight and accurate high-throughput phenotyping system for crop monitoring.

Abstract

LITERAL is a lightweight, portable high-throughput phenotyping tool. It meets the need for low-cost, easyto-use yet accurate measuring equipment for monitoring small plot trials or a network of agricultural plots. In practical terms, it integrates a set of sensors, including three high-resolution cameras, connected to an acquisition box that triggers acquisitions, stores data, and communicates with a tablet PC that enables measurement scenarios to be defined via a user-friendly graphic interface. The measurement scenario describes the configuration of each sensor, the test plan, and the number of measurements in each plot. This makes it easy to use in the field and ensures that each image is correctly referenced. Once downloaded, the data are analyzed by a modular processing chain, implementing generic processing algorithms: semantic segmentation, object detection by deep learning, colorimetric analysis, stereovision. These algorithms can be parameterized by culture to achieve high precision. The quality of the images acquired, and the many possible configurations mean that LITERAL can be used for a wide range of uses: monitoring the growth of field crops and trees, characterizing mixed crops, quantifying the symptoms of leaf diseases, measuring the density of plants or fruits, etc. Ergonomic and scalable, LITERAL has been developed as part of a CASDAR project led by ARVALIS and involving INRAE 1 , GEVES 2 , Terres Inovia 3 , ITB 4 , CTIFL 5 and HIPHEN. It is currently used by technical teams in France, Portugal, USA and Australia. Wider distribution is planned from 2024.

Code and data availability

The paper describes the LITERAL phenotyping system and its processing chains, but explicitly states that not all supporting data are available due to third-party involvement, and no public code, dataset, image, or model repository with an authors' URL is provided. Referenced resources (VegAnn, SegVeg, Global Wheat) are

No evidence-backed public reproduction asset is currently recorded.