摘要
以植被指数和红边参数为模型因子,利用多元线性回归(MLR),构建冬小麦不同生育期的生物量估算模型,从而有效和更好地监测冬小麦的长势情况,为精准农业中作物的快速监测提供技术手段。首先分析植被指数(VI)和红边参数(REPS)与冬小麦生物量的相关性,然后运用MLR分别建立模型MLR+VI、MLR+REPS和MLR+VI+REPS,最后将优选的冬小麦生物量估算模型应用于无人机高光谱影像中,验证模型的可行性。结果表明,利用单个植被指数或红边参数构建的估算模型在孕穗期、开花期和灌浆期估算精度最高的植被指数分别是归一化植被指数(NDVI)、简单比值指数(SR)和增强型土壤调节植被指数(MSAVI),精度最高的红边参数分别为红边振幅/最小振幅、红边振幅和红边振幅;通过MLR分别以植被指数、红边参数和植被指数结合红边参数为因子构建的模型MLR+VI、MLR+REPS与MLR+VI+REPS效果优于单个植被指数或红边参数建立的模型,3种模型在不同生育期的验证结果也较好,其中MLR+VI+REPS模型精度最高,模型决定系数(R 2)、标准均方根误差(NRMSE)分别为0.7832与12.13%。
Using vegetation index(VI)and red edge parameter(REPS)as model factors,multivariate linear regression(MLR)was used to construct a biomass estimation model for winter wheat in different growth periods,to effectively and better monitor the growth of winter wheat and provide technical means for rapid monitoring of crops in precision agriculture.The correlation of VI and REPS with biomass of winter wheat was analyzed first.Then MLR+VI model,MLR+REPS model and MLR+VI+REPS model were constructed by MLR respectively.Finally,the optimized model for estimation of biomass in winter wheat was applied in hyperspectral images taken by unmanned aerial vehicles to verify the feasibility of the models.The results showed that the vegetation indices with the highest estimation accuracy of the estimation model constructed by single vegetation index or red edge parameter in booting stage,flowering stage and filling stage were normalized difference vegetation index(NDVI),simple ratio index(SR)and modified soil-adjusted vegetation index(MSAVI)respectively,and the red edge parameters with the highest precision were red edge amplitude/minimum amplitude,red edge amplitude and red edge amplitude respectively.The effects of MLR+VI model,MLR+REPS model and MLR+VI+REPS model were better than the models constructed by single vegetation index or single red edge parameter.The verification results of the MLR+VI model,MLR+REPS model and MLR+VI+REPS model in different growth periods were also good,and the MLR+VI+REPS model showed the highest precision,the coefficient of determination(R 2)and normalized root mean square error(NRMSE)of the model were 0.7832 and 12.13%respectively.
作者
陶惠林
冯海宽
徐良骥
杨贵军
杨小冬
苗梦珂
刘明星
TAO Hui-lin;FENG Hai-kuan;XU Liang-ji;YANG Gui-jun;YANG Xiao-dong;MIAO Meng-ke;LIU Ming-xing(Key Laboratory of Quantitative Remote Sensing in Agriculture,Ministry of Agriculture and Rural Affairs,Beijing Research Center for Information Technology in Agriculture,Beijing 100097,China;School of Geodesy and Geomatics,Anhui University of Science and Technology,Huainan 232001,China;National Engineering Research Center for Information Technology in Agriculture,Beijing 100097,China;Beijing Engineering Research Center for Agricultural Internet of Things,Beijing 100097,China)
出处
《江苏农业学报》
CSCD
北大核心
2020年第5期1154-1162,共9页
Jiangsu Journal of Agricultural Sciences
基金
国家自然科学基金项目(41601346、41871333)。
关键词
无人机
高光谱
冬小麦
多元线性回归
植被指数
红边参数
unmanned aerial vehicle(UAV)
hyperspectral
winter wheat
multiple linear regression
vegetation index
red edge parameters
