| Biomass and LAI are important physiological and biochemical parameters of crops,which can effectively reflect the growth of crops.The level of yield is related to the quality of crops,the efficient monitoring of biomass and LAI,and accurate prediction of crop yield,which is important in agricultural production and at the same time has a strong guiding significance for agricultural managers.In this paper,digital images and hyperspectral data of three main growth stages of winter wheat were obtained by UAV.Winter wheat biomass estimation and LAI estimation based on UAV digital image were discussed mainly,winter wheat biomass and LAI estimation based on UAV hyperspectral data,winter wheat yield estimation based on UAV digital image and hyperspectral data,the main results and conclusions are as follows:(1)According to AIC and BIC,the best digital image index is selected,the flagging stage,the biomass modeling and the LAI estimation model are all the best modeling number 1,the R2 and NRMSE of the two models are 0.51,0.48,and 21.02%,26.16%,respectively,the model accuracy is low;the flowering stage,biomass estimation and LAI estimation are also the best number of modeling is 1,modeling R2 and NRMSE are 0.52,0.54 and 18.15%,23.65%,at this time the biomass estimation model has higher accuracy,and the LAI estimation model is less effective,but it is better than the flagging stage.During the filling stage,the optimal number of models is 4 and 3,and the R2 and NRMSE are 0.57,0.67,and 17.14%,30.69%.In summary,the biomass estimation model at the flowering stage is the best,followed by the filling stage,the flagging stage is the worst,the LAI estimation model at the flowering stage is the best,the flagging stage is the second,and the filling stage is the worst.(2)The spectral parameters are selected through correlation analysis to build a single parameter linear model for each growth stage.The three growth stage optimal parameters corresponding to biomass estimation and LAI estimation are NDVI,NDVI,OSAVI × SR,SR,NDVI × SR.Use MLR,PLSR,RF,and ANN to build biomass and LAI models.During the flagging stage,model biomass-MLR modeling R2,RMSE and NRMSE are 0.67,0.09 kg/m2,and 17.33%,respectively.LAI-MLR modeling R2,RMSE and NRMSE are 0.63,1.01 and 22.09%respectively;during flowering stage,biomass-MLR modeling R2,RMSE and NRMSE are 0.73,0.11kg/m2 and 13.59%respectively,and LAI-MLR modeling R2,RMSE and NRMSE are respectively 0.68,0.69 and 19.79%,the two models have the highest accuracy;during the filling stage,biomass-MLR and LAI-MLR still have the highest stability and the best estimation effect.(3)Based on the UAV digital image and imaging hyperspectral image data,the biomass and LAI estimation models constructed by PLSR are coupled with the univariate linear models established by the measured biomass and LAI and measured yields,respectively.The performance and fitting accuracy are the highest.Among them,the model modeling R2 based on the UAV digital image and biomass and LAI is 0.61 and 0.52,and the NRMSE is 15.44%and 15.57%,respectively;the estimation model based on the UAV imaging hyperspectral image and the biomass and LAI is established.R2 is 0.60 and 0.64,and NRMSE is 15.23%and 14.83%,respectively.Comparative analysis of yield models constructed from two different data sources,the model based on the hyperspectral image data of UAV imaging is more accurate and more suitable for estimating winter wheat yield.Figure[28]Table[23]parameters[99]... |
PDF Full Text Request