Leaf Area Index (LAI) Prediction Using Machine Learning and UAV Based Vegetation Indices

Abstract

As a critical indicator of plant growth and water use, accurately and promptly estimating leaf area index (LAI) is essential for improved crop production. However, measuring LAI using destructive methods is labor-intensive and time-consuming. The main objective of this study was to leverage vegetation indices (VIs) generated from unmanned aerial vehicle (UAV)-based images and machine learning (ML) algorithms for LAI estimation of green beans and sweet corn. The research experiments were conducted for three consecutive years during the winter (dry) seasons of 2020 through 2023 at the Tropical Research and Education Center (TREC), University of Florida, Homestead, Florida. The experiment consists of 32 plots with four irrigation treatments, i.e., 100% full irrigation (FI), 75%, 50%, and 25% FI, with four replications. Destructive leaf samples were collected by cutting plants from 30cm row length of two inner plot rows. The leaf area (LA) of each green beans and sweet corn sample was measured using the LI-3000C transparent belt conveyor. The plant height and width of these crops were also measured bi-weekly. The LAI of the plants was calculated using the plant density method along with measured LA. The DSSAT model calibrated in a previous study for simulating plant growth and yield was used to simulate the LAI of both crops. Moreover, a UAV-based RedEdge-MX sensor was employed throughout the seasons to collect high-resolution multispectral imageries that consist of five bands. Twelve VIs were generated from UAV-based multispectral images. The DSSAT model simulated LAI for both crops was validated against plant-estimated LAI, and results depicted a good correlation for green beans and reasonable for sweet corn. Furthermore, validated DSSAT LAI values were compared with the twelve VIs to build a relationship between LAI and VIs. Out of 12 indices, six VIs, i.e., EVI2, NDVI, NGRDI, NIRRENDVI, RENDVI, and SAVI showed a good agreement with LAI for both crops. Additionally, ML algorithms, i.e., random forest (RF), eXtreme gradient boosting (XGB), and light gradient boosting (LGB) models, were trained to predict the LAI of green beans and sweet corn using VIs as input features. The LGB, RF, and XGB models predicted LAI with acceptable accuracy, achieving r² values of 0.78, 0.90, and 0.90 with RMSE values of 0.43, 0.29, and 0.28 respectively for sweet corn and r² of 0.72, 0.79, and 0.80 and RMSE of 1.01, 0.86, and 0.85 respectively for green beans. Therefore, it is concluded that ML models can predict the LAI with good accuracy for green beans and sweet corn using UAV multispectral image-based VIs as input features.

Plant phenotyping relevance

UAVマルチスペクトル画像と機械学習を用いて作物のLAIを推定する手法が研究の中心であり、植物形質の取得・推定方法を技術的に評価している。

Code and data availability

公開状態または取得可能な本文経路を確認できませんでした。

No evidence-backed public reproduction asset is currently recorded.