Remote Sensing Inversion Of Pear Tree Biomass And Leaf Area Index In Southern Xinjiang

Korla Fragrant Pear is popular in the market because of its sweet and crisp taste,small and exquisite appearance and high nutritional value.Based on the UAV platform,it can quickly monitor the crop growth information and reduce the damage to the plant.In this study,from May to September 2021,the trunk pear garden of the 15 th company of the 9th regiment of alar,Xinjiang was taken as the experimental site.The visible light and multispectral remote sensing image data of the experimental site were obtained through the UAV platform.After image preprocessing,five vegetation indexes were extracted through band operation,and compared with the pear leaf area index and aboveground dry biomass data obtained by direct observation,weighing and image processing,The empirical model method and BP artificial neural network method are used to construct the inversion model respectively.Based on the prediction accuracy of the model,the two models were evaluated and analyzed,the models with high accuracy were selected,and the inversion maps of aboveground dry biomass and leaf area index of pear trees in germination stage,fruit setting stage,fruit expansion stage and maturity stage in the experimental area were drawn.The results show that:(1)In the empirical regression model for the whole growth period of pear tree based on five vegetation indexes and fragrant pear leaf area index,the prediction results of the index model of OSAVI and fragrant pear leaf area index are better,the determination coefficient R2 is 0.681,the root mean square error RMSE is 0.531,and the prediction accuracy is 83.32%.In the empirical model of the whole growth period of pear tree based on vegetation index and aboveground dry biomass of fragrant pear,the prediction result of the index model constructed by GNDVI and aboveground dry biomass of fragrant pear is better,R2 is 0.652,RMSE is 1.56 kg / plant,and the prediction accuracy is 76.85%.(2)In the BP neural network model for the whole growth period of pear tree based on five vegetation indexes,pear leaf area index and aboveground dry biomass,all vegetation indexes are mixed as the input layer,and the inversion effect is better.In the regression model constructed with leaf area index,the R2 of vi-bp model is 0.752,RMSE is 0.475,and the prediction accuracy is 85.29%;In the regression model constructed with pear aboveground dry biomass,the R2 of vi-bp model is 0.770,RMSE is 0.913 kg / plant,and the prediction accuracy is 81.36%.In BP neural network,the mixed vegetation index as the input layer contains more bands than a single vegetation index,which can improve the inversion accuracy.(3)The accuracy of the inversion results constructed by BP neural network and empirical model is evaluated.The inversion accuracy of BP neural network in retrieving pear leaf area and aboveground dry biomass is higher than that of empirical model.In terms of leaf area index inversion,the inversion accuracy of BP neural network optimal model is 1.97% higher than that of empirical model.In terms of inversion of aboveground dry biomass of pear trees,the inversion accuracy of the optimal model of BP neural network is4.51% higher than that of the empirical model...