Monitoring Of Winter Wheat Growth Index Based On UAV Multi-spectral Remote Sensing

As one of the three major food crops,real-time monitoring of the growth of wheat is crucial to improving the quality and yield of wheat.UAV remote sensing,as a precision agriculture application technology for accurate and efficient monitoring of crop growth,is a current hot spot in agricultural information technology research.In this study,based on the multispectral images of winter wheat canopy collected by the UAV multispectral remote sensing platform,ten representative vegetation indices were constructed and analysed in combination with ground-based measured growth parameters,and different methods were investigated to invert the leaf area index and above-ground biomass of winter wheat and map the spatial and temporal distribution of leaf area index and above-ground biomass.The main research elements are as follows:(1)Leaf area index and above-ground biomass of winter wheat were both weakly correlated with red-edge band reflectance,while leaf area index was highly significantly correlated with red and NIR band reflectance with a correlation coefficient of 0.65 or more,and above-ground biomass was highly significantly correlated with red and NIR band reflectance with a correlation coefficient of 0.81 or more.In the correlation analysis of leaf area index,above-ground biomass and vegetation index of wheat,it was found that NDVI,RVI,DVI,GNDVI and NDRE correlated well with leaf area index and above-ground biomass,with absolute values of correlation coefficients greater than 0.60.(2)Research on leaf area index estimation based on multispectral images from drones.Three modelling methods,namely stepwise regression,support vector regression and random forest regression,were used to construct and evaluate the inverse model of the leaf area index of winter wheat.The support vector regression model performed best in the inversion of leaf area index,with R~2 of 0.78 and 0.80 in the modelling and validation sets,respectively,which were higher than those of the stepwise regression model(0.60 and 0.73)and the random forest model(0.65 and 0.71),and the corresponding RMSE were0.32 and 0.30,which were lower than those of the stepwise regression model(0.43 and 0.34)and the random forest model(0.40 and 0.36).(3)Research on the estimation of above-ground biomass based on multispectral images from drones.Three modelling methods,namely stepwise regression,support vector regression and random forest regression,were used to construct an inversion model of above-ground biomass of winter wheat and to evaluate the model.The random forest regression model performed best in the inversion of above-ground biomass,with R~2of 0.78 and 0.84 in the modelling and validation sets,respectively,which were higher than those of the stepwise regression model(0.76 and 0.81)and the random forest regression model(0.76and 0.82),corresponding to RMSE of 115.49 g/m~2 and 96.00 g/m~2,which were lower than those of the stepwise regression model(124.75 g/m~2 and 104.07 g/m~2)and the random forest model(120.34 g/m~2,101.81 g/m~2).(4)The spatial and temporal distribution of leaf area index and above-ground biomass were mapped based on the optimal model.Temporally,the leaf area index and above-ground biomass showed a gradual increase with the advancement of wheat fertility,while the growth of leaf area index slowed down at the filling stage and above-ground biomass continued to increase.Spatially,the leaf area index and above-ground biomass behaved more or less the same under different fertiliser treatments,and the growth under different fertiliser treatments could be inferred more accurately by the drone.