Research On Hyperspecrtal Characteristic Of Cotton And Remote Sensing Inversion About Cotton Agronomic Parameters

Cotton is one of the primary economic crops in our country,it is the textile and fine chemical raw material,and the important strategic material.In Weibei dry highland,ecological environment is fragile,water resources are scarcity and uneven distribution of time and space,the soil fertility is low.Need to change the traditional agricultural production mode,implement precision agriculture,to improve the efficiency of agricultural production and protect the local ecological environment.The agricultural parameters inversion model with hyperspectral remote sensing can realize the dynamic monitoring of crop growth,thus providing a basis for the implementation of precision agriculture.This research took cotton in Weibei dry highland as study object,and obtained the hyperspectral data and agronomic parameters from field experiment on ground and UAV scales.To study cotton spectral characteristics of the agronomic parameters,the correlation of agronomic parameters and canopy spectral reflectance,the correlation of agronomic parameters and first derivative spectral,and the correlation of agronomic parameters and vegetation indexes.To establish cotton agronomic parameters estimation models by single factor polynomial linear regression method、multiple stepwise regression(MSR)method and support vector machine regression(SVM)method based on vegetation indexes,and comparing the accuracy of each estimation model to get the best estimation model for the agricultural parameters of cotton.At last,use the best estimation model for the agricultural parameters of cotton to remote sensing inversion of UAV images.The main conclusions and achievements are as follows:(1)The content of chlorophyll in cotton first increased and then decreased in different growth stages.The chlorophyll content was increased from seedling stage to flowering and boll-setting stage,and the chlorophyll content was decreased from flowering and boll-setting stage to boll opening stage.The spectral reflectance of the cotton canopy was decreased with the SPAD values increased in the visible band,and the higher SPAD values in the NIR band,the higher spectral reflectance of the cotton canopy.In the red edge band,the red edge feature apperared the phenomenon of"redshift".The chlorophyll content in all growth stages is significant negative correlation with the original spectral reflectivity in the range of 520-600nm and 690-720nm,and is significant positive correlation in the near infrared band of 760-1000nm.The SVM regression method can effectively improve cotton hyperspectral inversion precision of chlorophyll content,and it can be used as the preferred method of cotton chlorophyll conten hyperspectral remote sensing inversion.Verification accuracy of SVM regression model SPAD-SVM_b with more vegetation indexes is better than all other models,which is the best model for hyperspectral inversion of chlorophyll content in this study.(2)The content of anthocyanin in cotton first decreased and then increased in different growth stages.The anthocyanin content was decreased from seedling stage to flowering and boll-setting stage,and the anthocyanin content was increased from flowering and boll-setting stage to boll opening stage.The spectral reflectance of the cotton canopy was increased with the Anth values increased in the visible band,and the higher Anth values in NIR band,the lower spectral reflectance of the cotton canopy.In the red edge band,the red edge feature apperared the phenomenon of"blue shift".The anthocyanin content in all growth stages is significant positive correlation with the original spectral reflectivity in the range of 520-600nm and 700-720nm,and is significant negative correlation in the near infrared band of 780-1000nm.The SVM regression method can effectively improve cotton hyperspectral inversion precision of anthocyanin content,and it can be used as the preferred method of cotton anthocyanin conten hyperspectral remote sensing inversion.Verification accuracy of SVM regression model Anth-SVM_b with more vegetation indexes is better than all other models,which is the best model for hyperspectral inversion of anthocyanin content in this study.(3)The spectral reflectance of the cotton canopy was decreased with the LAI increased in the visible band,and the higher LAI in the near infrared band,the higher spectral reflectance of the cotton canopy.In the red edge band,the red edge feature apperared the phenomenon of"redshift".The cotton LAI of cotton is significant negative correlation with the original spectral reflectivity in the range of 510-730nm,and is significant positive correlation in the near infrared band of 760-1000nm.The SVM regression method can effectively improve cotton hyperspectral inversion precision of cotton LAI,and it can be used as the preferred method of cotton LAI hyperspectral remote sensing inversion.Verification accuracy of SVM regression model LAI-SVM_b with more vegetation indexes is better than all other models,which is the best model for hyperspectral inversion of cotton LAI in this study.(4)Using the best estimation model of SPAD,Anth and LAI of cotton canopy to map the hyperspectral remote sensing image of UAV.To obtain the spatial distribution map of cotton canopy SPAD,Anth and LAI,and the inversion results are basically correct.Therefore,the estimation models of cotton agronomic parameters in this study has applicability in the high spectral inversion of UAV scale.