Remote Sensing Model Of Physiological Parameters Of Summer Maize At Different Growth Stages Based On Machine Learning Algorithms

Chlorophyll content and leaf water content are important indicators of corn growth status.In this study,the summer maize widely planted in the northwestern region which was used as the research object.The non-imaging hyperspectral data of summer maize in different growth stages were determined by using the spectral spectroradiometer.The chlorophyll content and leaf water content data were obtained simultaneously,and the hyperspectral data of different growth stages were analyzed.Based on the correlation between remote sensing data and the inversion model based on characteristic wavelength,vegetation index and hyperspectral characteristic parameters.The main conclusions are as follows:1.Analysis of chlorophyll content and leaf water content:The chlorophyll content increased from the jointing stage to the milky stage.As the plant ageed,the maturity period began to decrease.The spectral reflectance of each period was significantly different in the“green peak”region;the mean value of leaf water content increased continuously,but the growth rate increased from The milk ripening period begins to slow down.2.Correlation analysis:The correlation between the chlorophyll content and the spectrum of summer maize was the highest correlation between the original spectral reflectance and the chlorophyll content at547 nm,342 nm,719 nm and 343 nm from the jointing stage,the tasseling stage,the milky stage to the ripening stage.The maximum correlation coefficient was-0.170,-0.293,-0.476,-0.196;the first-order differential spectra and chlorophyll content were most correlated at the wavelengths of 964 nm,835 nm,707 nm,and 910 nm,and the maximum correlation coefficients were 0.482,0.358,-0.499,and 0.269,respectively;VOG1,MTCI,and TVI are the optimal vegetation indices for each growth period,and the correlation coefficients with chlorophyllcontentare0.284,-0.279,0.507,-0.127;(SDr-SDb)/(SDr+SDb),(SDr-SDb))/(SDr+SDb),SDr/SDb,and SDy are the optimal hyperspectral remote sensing parameters for each growth period,and the correlation coefficients with chlorophyll content are 0.284,0.285,0.479,and 0.133.The water content of summer maize leaves had the highest correlation with the original growth spectra at 450nm,367nm,353nm and 353nm,respectively.The maximum correlation coefficients were-0.580,-0.266,0.337,-0.165,respectively.The differential spectra have the largest correlation at 1654nm,960nm,1072nm and 1070nm.The maximum correlation coefficients are-0.588,-0.478,-0.450,-0.388.NDVI,NDVI,WI and VOG1 are the best vegetation indices for each period.The parameters of the hyperspectral remote sensing parameters SDy,Ry,Ry,and Db were the most correlated with the leaf water content of each period,and the correlation coefficients were 0.334,0.253,-0.342,and 0.308,respectively.3.Optimal estimation model:The chlorophyll content inversion,the univariate optimal estimation model is the KNN model constructed with the first-order differential eigenwaves in the milky stage.The model modeling and verification R~2 are 0.366 and 0.330,respectively;the optimal multivariate regression model is the milky maturity vegetation index.The constructed XGBoost model has R~2 of 0.525 and 0.821 for modeling and verification.Inversion of leaf water content,the univariate optimal estimation model is the XGBoost algorithm model established by SDr/SDb parameters during the maturity period.The modeling and verification R~2 are 0.741 and 0.285 respectively.The multivariate optimal estimation model is constructed for the jointing vegetation index.The XGBoost multiple regression model,modeling and verification R2 were 0.264,0.298,respectively.