Estimation of Cotton Leaf Area Index under Verticillium Wilt Stress ： Based on UAV Multispectral & Optimization Algorithm

Abstract

Abstract Verticillium wilt is one of the main factors hindering cotton yield increase, the more severe the disease, the more severe the yield loss. In order to achieve early detection and prevention of Verticillium wilt, this study investigated cotton plants naturally affected by Verticillium wilt in the field. We analyzed the cotton canopy multispectral data under the stress of Verticillium wilt and the leaf area index (LAI) data collected from the ground. Due to the low prediction accu-racy of traditional empirical models, this paper selected three different back propagation BP neural network models with 5, 10, and 15 hidden layer nodes (hln) and optimized them using four intelligent swarm algorithms: genetic algorithm (GA), particle swarm optimization (PSO), spotted hyena algorithm (SHO), and improved spotted hyena algorithm (ISHO) to estimate the cotton LAI under the stress of Verticillium wilt. Finally, BP algorithm with different node number of hidden layer was optimized by comparing four intelligent swarm optimization al-gorithms to estimate LAI of cotton under verticillium wilt stress. The findings indicated that GA-BP (hln15) was the best in the GA-BP algorithm; In the PSO-BP algorithm, PSO-BP (hln10) performed best; In the SHO-BP model, SHO-BP (hln5) had the highest model performance; In the ISHO-BP model, ISHO-BP (hln5) is the best, with a determination coefficient(R2), root mean square error(RMSE), and prediction accuracy(PA) of 0.952, 0.235, and 89.30%, respec-tively; The estimation results of ISHO-BP (hln5) model can better represent the distribution of Verticillium wilt plants than GA-BP (hln15), PSO-BP (hln10), and SHO-BP (hln5), and its estimation error range is (-0.8,0.5). Therefore, the application of ISHO-BP (hln5) model pro-vides a new method for estimating cotton LAI under pest and disease stress, and also provides technical support to meet the diversified needs of cotton farmers to increase their income and national economic development.

Plant phenotyping relevance

UAVマルチスペクトルデータから綿のLAIを推定するモデルを比較・最適化しており、植物形質の取得・推定手法が研究の中心である。

Code and data availability

The supplied blocks describe UAV multispectral imagery, ground LAI measurements (150 samples), and MATLAB-based GA/PSO/SHO/ISHO-BP modeling, but contain no data availability statement, no public dataset or code deposit, and no author-provided URLs. No paper-specific public asset is identified.

No evidence-backed public reproduction asset is currently recorded.