Research On Cotton Chlorophyll Inversion And Yield Estimation Model Based On UAV Remote Sensing

Timely and accurate monitoring of the dynamic information of crop growth and yield is of great significance to the refined management of crop fields.In this study,the unmanned aerial vehicle multispectral remote sensing platform was used to collect the spectral images of the leaf canopy in four key growth stages of cotton bud stage,flower bud stage,flower boll stage and boll opening stage.Combined with the chlorophyll value and final yield of cotton in each growth stage,statistical Model analysis,established cotton chlorophyll inversion model and cotton yield prediction model respectively,and compared different modeling methods(univariate linear regression UR,multivariate stepwise regression SMLR,principal component regression model PCA)in the cotton chlorophyll inversion and growth period.The applicability of the production estimation model was obtained,and the optimal chlorophyll inversion model and the optimal production estimation model were obtained.The results of the study are as follows:(1)The chlorophyll value of cotton canopy leaves was significantly or extremely significantly correlated with most spectral parameters.Red light and blue light were significantly negatively correlated with chlorophyll value throughout the growth period.Chlorophyll values showed a significant positive correlation.(2)Three regression models were used to construct a chlorophyll value inversion model based on 15kinds of leaf canopy spectral parameters,and it was found that there was a significant or extremely significant correlation between the predicted value of chlorophyll content and the measured value.The constructed model has the highest accuracy and is the best model for inversion of cotton chlorophyll value,and its R~2,RMSE and n RMSE are 0.767,0.92 and 1.55%.(3)Through the significance test,the correlation between each spectral parameter and yield at different growth stages showed significant or extremely significant correlation,and the correlation between chlorophyll content and yield at different growth stages showed the following order:flower boll stage>bud stage>flower bud stage>Boll-opening period;the correlation between normalized vegetation index NDVI and yield was relatively high in different growth stages as a whole,and R~2 was 0.605,0.741,0.835 and 0.696in different growth stages,respectively.(4)Among the cotton yield estimation models constructed by using the three regression models,the SMLR model constructed by combining chlorophyll value and spectral parameters as independent variables has the best yield estimation effect,with R~2,RMSE and n RMSE of 0.923,67.54 kg/hm~2 and 1.92%,in different growth stages,the models constructed by combining spectral parameters and chlorophyll values as independent variables in the SMLR and PCA models had the highest accuracy,and the predictive ability of the models was relatively stable.