无人机多光谱遥感监测稻瘟病的分级诊断模型研究OA

Rice blast is one of the most harmful diseases of rice cultivation,and it significantly reduces yields.The use of unmanned aerial vehicle(UAV)multispectral imaging technology holds considerable promise in terms of rapid and large-scale surveillance.In this study,a stratified diagnostic model for rice blast was developed using canopy multispectral information continuously collected by unmanned aerial vehicles(UAVs).The accuracy,precision,F1-score and recell of the stacking model were as high as 89.71％,89.29％,89.42％and 89.8％,respectively.During the study,12 vegetation indices(VIs)were sorted by correlation characteristics using a random forest to screen for the best combination of vegetation indices(VIs).The results showed that the three vegetation indices(VIs)of NDRE,NDVI and RVI were the most effective for the stratified diagnosis of rice blast throughout the growth period.The accuracy,precision,F1-score and recell of the stacking model were as high as 91.18％,90.34％,90.67％and 91.07％,respectively.This increased by 1.47％,1.05％,1.25％and 1.27％respectively.This study not only verified the validity of hierarchical diagnosis but also provided strong support for precise treatment,assisted fertilization and drug application,reduced environmental pollution and improved yield.