Research On Inversion Method Of Chlorophyll Content In Rice Canopy Leaves Based On UAV Hyperspectral Remote Sensing

Rice is the main food crop in my country,and the precise management of its production process is of great significance to improve rice yield and ensure food security.The chlorophyll content of rice is an important monitoring factor for precision agriculture,and its content is of great significance to the health status of rice,agricultural irrigation and fertilization regulation.With the development and application of imaging hyperspectral technology and UAV remote sensing,it is possible to use quantitative remote sensing technology to accurately obtain information on chlorophyll content in rice.In order to explore effective rice chlorophyll content spectral features（Chlorophyll specific spectral features,CSSFs）and inversion modeling methods,and solve problems such as unmanned aerial vehicle remote sensing monitoring of rice chlorophyll content.Unmanned aerial vehicle rice canopy hyperspectral data and ground sample data to carry out experimental research,the main work includes:（1）The spectral response characteristics of chlorophyll content in rice were analyzed,and it was found that the change of chlorophyll content had a significant impact on the spectral response.The lower the chlorophyll content,the higher the reflectance in the range of523～665nm and 700-800nm;the increase of chlorophyll content will make the spectrum in the700-800nm range.A red shift occurs;in the first-order differential spectrum,the higher the chlorophyll content,the lower the derivative value in the 490～560nm and 686～710nm bands;An increase will redshift the first-order differential spectrum.（2）A total of 86 CSSFs were obtained by applying spectral response analysis,various vegetation index analysis,red edge location extraction,and Regularized Neighborhood Component Analysis（RNCA）,and further designed the features of significance test and collinearity reduction.With the optimized method,13 CSSFs were finally obtained,respectively:4 spectral response features:₂)、₀)、₂)and;6 vegetation index features:carotene vegetation index（PRI,PNIR×CRI700）,lutein vegetation index（PRIm1）,chlorophyll vegetation index（VOG2,CTRI1,m NDI）;1 red edge location feature:REP_LE;2 regular neighbor component analysis features:R₇₁₅,R₇₀₇.In the RNCA feature extraction algorithm,when the regularization parameter is 0.306 and the loss function is the root mean square error function,the feature selection performance is optimal.（3）The chlorophyll content inversion model was constructed by using particle swarm-extreme learning machine,Gaussian process regression,support vector machine regression,multiple linear regression and other methods.0.831,an increase of 7.922%compared to the traditional extreme learning machine model;followed by the Gaussian process regression model,the exponential type performed the best among the three kernel functions,with R²reaching 0.786,and the square exponential kernel and quadratic rational kernel R² were 0.775and 0.744,respectively.The results of this study can provide theoretical support for high-throughput crop monitoring based on UAV platform.