Inversion Of Apple Tree Canopy Nitrogen Content Based On Ground And Aerial Integrated Remote Sensing Data

The rapid monitoring of vegetation nitrogen content has always been a hot issue in research,and how to improve the inversion accuracy of nitrogen content has been a problem At present,ground remote sensing and Unmanned Aerial Vehicle(UAV)remote sensing is widely used in the field of precision agriculture.Although the spectral reflectance obtained by the ground spectrometer is relatively accurate,there are shortcomings of discrete point measurement.UAV remote sensing is convenient,fast and can perform continuous area monitoring,but the spectral reflectance accuracy is lower than the ground spectrometer Integrating the ground and UAV remote sensing data can provide technical support for regional accurate remote sensing inversionQixia City,located in Jiaodong Peninsula and Mengyin County,located in Tai Yi Mountain are the main apple production district in Shandong Province.So we chose them as the research areas.We take the Red Fuji apple trees in the new shoots stage as the research object.DJ M600 Pro low-altitude UAV equipped with a multispectral camera was used to collect multispectral images of the apple canopy,and the ground spectrometer ASD FieldSpec4 was used to collect canopy and bare soil hyperspectral data.We used spectral reflectance data of UAV multispectral remote sensing images to construct spectral characteristic parameters,analyze the correlation between the nitrogen content of apple tree canopy and them,and select sensitive spectral characteristic parameters to establish apple tree canopy nitrogen content estimation models based on spectral characteristic parameters.Then,we used the ground canopy and bare soil endmember spectral data,and used a linear mixed pixel model to perform mixed pixel decomposition on the UAV remote sensing image.Then the canopy abundance parameters based on ground and aerial integrated remote sensing data were obtained.Analyze correlation between canopy abundance parameters and nitrogen content.Then we explore the effect of introducing abundance parameters into the nitrogen content inversion model on the accuracy of model inversion so as to establish the optimal nitrogen content inversion model.Finally,the apple tree canopy was extracted from the UAV multispectral image by supervised classification,and the nitrogen content spatial inversion of the apple tree canopy region from the UAV remote sensing image was performed by the optimal nitrogen content model.The main results are as follows(1)Established and chose spectral characteristic parameters related to canopy nitrogen content18 types of vegetative index spectral characteristic DVI,RVI(E,N),RVI(N,G),RVI(N,R),NDVI,RDVI,SAVI,OS AVI,NRI,GNDVI,NDCI,TCARI,TVI,MCARI,NLI,MNLI,MSAVI2 and MSR,constructed by the original spectral reflectance,were correlated with nitrogen content.We chose the spectral characteristic parameters with absolute value of correlation coefficient larger than 0.5.Therefore,GNDVI and RVI(N,G)were screened out35 types of the non-vegetative index spectral characteristic parameters G,R,E,N,lnG,lnR,lnE,lnN,1/G,1/R,1/E,1/N,(?),(?),(?),(?),G2,G3,GXR,GXE,GXN,GXRXE,G×R×N,G×E×N,R×E×N,G×R×E×N,R2,R3,RX E,R×N,E2,E3,EXN,N2,N3 and nitrogen content were taken for correlation analysis.The six non-vegetation index spectral characteristic parameters were selected,which are G,G×R×E,G×R×E×N,G×E,GXR and RXE(2)Canopy abundance parameters of apple tree based on ground and aerial integrated remote sensing dataThe correlation coefficient between the canopy abundance value and nitrogen content of Qixia Boshida and Mengyin Yedian apple orchard is 0.688 and 0.741,respectively.They can be used as the independent variable for establishing and testing the canopy nitrogen content inversion model.(3)Established and tested the model for estimating canopy apple tree nitrogen contentLR,PLSR,SVR and RFR models were established by G,GNDVI,RVI(N,G),G×R,G×E,R×E,G×R×E,G×R×E×N and canopy abundance parameter of the calibration set as independent variables and nitrogen content as dependent variable in Qixia Boshida apple orchard.We test the model with validation sets.The R2,RMSE and RPD of the regression model of the SVR model verified by the Qixia Boshida apple orchard validation set are 0.820,0.079 and 2.113,respectively.The R2,RMSE and RPD of the SVR model further verified by Mengyin Yedian apple orchard validation set are 0.848,0.080 and 2.189,respectively.SVR model is suitable for nitrogen content estimation of apple tree canopy in the new shoots stage.(4)Determined the best inversion model of apple tree canopy nitrogen content and completed spatial inversionThe best inversion model of apple canopy nitrogen content is obtained based on the SVR model of canopy abundance and spectral characteristic parameters,which was used to spatially estimate the nitrogen content of the canopy regionThe ground-aerial integrated remote sensing data was used to estimate the nitrogen content of apple canopy,which provides theoretical basis and technical support for rapid detection of regional fruit tree nutrients.