| 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 r2, RMSE, and nRMSE were 0.866, 0.95 g·kg-1, and 6.23%, respectively, and the model validation r2, 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. |
基本信息
