基于植被指数融合的无人机冬小麦LNC反演  

Research on UAV Winter Wheat LNC Inversion Based on Fusion Vegetation Index

作　　者：愿彬彬 汪洋 武红旗[1,3] 康镱梁 谷海斌 骆俊腾 王帅帅 YUAN Binbin;WANG Yang;WU Hongqi;KANG Yiliang;GU Haibin;LUO Junteng;WANG Shuaishuai(College of Resources and Environment,Xinjiang Agricultural University,Urumqi,Xinjiang 830052,China;College of Grass Industry,Xinjiang Agricultural University,Urumqi,Xinjiang 830052,China;Laboratory of Soil and Plant Ecological Processes,Xinjiang,Urumqi,Xinjiang 830052,China)

机构地区：[1]新疆农业大学资源与环境学院,新疆乌鲁木齐830052 [2]新疆农业大学草业学院,新疆乌鲁木齐830052 [3]新疆土壤与植物生态过程实验室,新疆乌鲁木齐830052

出　　处：《麦类作物学报》2024年第8期1063-1073,共11页Journal of Triticeae Crops

基　　金：国家自然科学基金项目(31560340)。

摘　　要：为了解无人机遥感平台用于快速、准确地监测冬小麦叶片氮含量(LNC)中的可行性,利用无人机遥感平台获取新疆喀什地区新疆农业科学院小麦育种基地冬小麦冠层光谱图像,分析和筛选可见光植被指数、多光谱植被指数与LNC的相关性,建立融合植被指数,比较多元线性回归(MLR)、逐步线性回归(SMLR)、随机森林回归(RF)在冬小麦各生育时期对叶片氮含量的适用性,筛选最优冬小麦叶片氮素含量估测模型。结果表明,小麦LNC与可见光植被指数(ExR、IKAW、VARI)、多光谱植被指数(RVI、RDVI、MSR、NDRE、RERDVI)、融合植被指数(ExR×RERDVI、IKAW×RERDVI和VARI×RERDVI)具有较高相关性,遥感监测效果在抽穗期最佳,灌浆期次之,成熟期最差。以融合植被指数作为自变量,采用随机森林回归模型构建的LNC估测模型在抽穗期的预测精度最佳,建模r^(2)、RMSE和nRMSE分别为0.866、0.95 g·kg^(-1)和6.23%,模型验证r^(2)、RMSE和nRMSE分别为0.71、1.61 g·kg^(-1)和10.83%。这说明基于无人机遥感平台利用融合植被指数能够实现对冬小麦LNC的快速、准确估测。To understand the feasibility of UAV remote sensing platform for rapid and accurate monitoring of nitrogen content(LNC)in winter wheat leaves,the UAV remote sensing platform was used to obtain winter wheat canopy spectral images from the wheat breeding station of Xinjiang Academy of Agricultural Sciences in Kashgar,Xinjiang.The correlation between visible light vegetation index,multispectral vegetation index,and LNC were analyzed and screened,and a fusion vegetation index was established.The applicability of multiple linear regression(MLR),stepwise linear regression(SMLR),and random forest regression(RF)to leaf nitrogen content at different growth stages of winter wheat were compared,and the optimal model for estimating leaf nitrogen content in winter wheat was selected.The results showed that wheat LNC had a high correlation with visible light vegetation indices(ExR,IKAW,VARI),multispectral vegetation indices(RVI,RDVI,MSR,NDRE,RERDVI),and fusion vegetation indices(ExR×RERDVI,IKAW×RERDVI,and VARI×RERDVI).Remote sensing monitoring was the best at heading stage,followed by grain filling stage,and worst at maturity stage.The LNC estimation model constructed using a random forest regression model with the fusion vegetation index as the independent variable showed the best prediction accuracy at heading stage,and the modeling r^(2),RMSE,and nRMSE were 0.866,0.95 g·kg^(-1),and 6.23%,respectively,and the model validation r^(2),RMSE and nRMSE were 0.71%,1.61 g·kg^(-1),and 10.83%,respectively.This indicates that the remote sensing platform can achieve rapid and accurate estimation of winter wheat LNC using the fusion vegetation index.

关 键 词：无人机 冬小麦 叶片氮含量 植被指数 可见光 多光谱 

分 类 号：S512.1[农业科学—作物学] S311