Research On Inversion Model Of Leaf Area Index And Nutrient Of Litchi Based On Remote Sensing Data

Litchi is a typical tropical fruit in south China,mainly planted in Guangdong,Guangxi,Fujian,Hainan and other regions.Litchi planting is one of the main ways for farmers to increase their economy in south China,and its yield is very important.In order to improve the yield of Litchi and obtain the information of crop nutrition and growth,it is very important to put forward scientific and reasonable fertilization management.In order to obtain the growth information of Litchi quickly and accurately,the monitoring models of hyperspectral and multi-spectral growth information of unmanned aerial vehicles(uavs)are established respectively in this paper,taking the typical Litchi plantation in Guangdong Province as the research object,the model performances of ground hyperspectral and multispectral are compared.In order to improve the precision of the model,the nutrient distribution characteristics of the canopy were analyzed,and the nutrient spectral model was established by using the method of Multispectral image and regression statistical analysis,the effect of multi-angle on the nutrient retrieval model is discussed,which provides a theoretical basis for the optimization of multi-spectral nutrient monitoring.The main conclusions are as follows:1)Based on the measured canopy reflection spectrum of litchi in the autumn maturity stage,the correlation between the canopy reflection spectrum and leaf area index,leaf nitrogen,phosphorus,and potassium was studied.The results showed that after the calculation of first derivative(DR)and the inverse logarithmic first derivative(Dlg1/R),the canopy reflectance spectra were analyzed.Nitrogen has a significant correlation with Dlg1/R spectra near 635 nm and 1025 nm.There is a significant correlation between phosphorus and DR spectra value near 555 nm.The Dlg1/R spectra near 545 nm and 635 nm showed significant correlation.Potassium has a significant correlation with Dlg1/R spectral values near 45,775 and 1035 nm.2)Correlation analysis of canopy reflectance spectra and leaf nutrients during Litchi fruit development showed that the model determination coefficients of each index R~2were greater than 0.65 after the canopy reflectance spectra were optimized by DR,Dlog1/R and band ratio,which indicated that it was feasible to estimate leaf area index and Leaf N,P,k content by using these spectral variables.The R~2of each index is less than 0.53 in autumn mature period,which indicates that leaf area index and N,P,K estimate in autumn mature period.3)There are spatial distribution characteristics of litchi canopy nutrients,and there is a difference in the correlation between leaf nutrients in the upper,middle and lower canopy and vegetation index,and there are also differences in the models established.When inverting crop nitrogen from vertically observed remote sensing data,it was found that the accuracy of nitrogen inversion in the upper canopy was higher,with the determination coefficient R~2reaching 0.79,the residual predictive deviation(RPD)is 1.5,and the middle potassium content model determining coefficient reaching 0.74.4)Multi-angle and multi-spectral image data sets can improve the accuracy of nutrient inversion.During fruit development,the coefficient of determination of the optimal PLSR model for nutrient inversion(R~2>0.57),and the coefficient of determination of the optimal PLSR model for flower bud differentiation(R~2>0.59).5)In the similar process of fruit growth and development,the precision of the optimal multi-band univariate regression monitoring model of ground hyperspectral data(N:R~2=0.92,P:R~2=0.98、K:R~2=0.77)is better than that of the optimal vegetation index monitoring model of UAV multi-spectral data(N:R~2=0.71,P:R~2=0.72、K:R~2=0.74),but the data acquisition of UAV multi-spectral data is more efficient and cheaper than that of ground hyperspectral data.