The Estimate Severity Level of Cotton Verticillium Wilt Using New Multi-spectra of UAV Comprehensive Monitoring Disease Index

Verticillium wilt is one of the important diseases of cotton. In severe cases, it can lead to a reduction in cotton production or even no harvest. The severity level of Verticillium wilt of cotton were estimated by using the VAV (Unmanned Aerial Vehicle) multi-spectral Comprehensive monitoring disease index model, it can provide reference for estimating crop diseases. Dates were acquired from UAV multi-spectral image data and ground survey, correlation coefficient method and the best exponential factor method were used to select the best vegetation index and best band combination, A new UAV multi-spectral Comprehensive monitoring disease index was created, and four kinds of regression models which were on the new disease Comprehensive index were established to estimate cotton disease severity. The results show that cotton plants with different severity of Verticillium wilt have different spectral characteristics in the near-infrared and visible light bands. Namely, the disease severity increasing, the spectral reflectance of cotton canopy of disease was show an upward trend in the visible light band (470nm~656nm), however, a downward trend in the near-infrared band (710nm~950nm). DVI, B3-B5-B8 were the best vegetation index and best band combination for identify cotton disease by UAV multi-spectral data, respectively. A new index, that was B-RBDVI (RB (3-5-8) +DVI), named UAV multispectral Comprehensive monitoring disease index was created through combination DVI and B3-B5-B8. With this Comprehensive index as the independent variable, the regression models of multiple linear, partial least squares, principal component analysis and support vector machine were established to estimate the cotton disease severity. The R2 of the prediction set of the four regression estimation models were 0.879 -0.912, with a difference of 0.065-0.079 in RMSE. The R2 of the validation set is between 0.88-0.89, with a difference of 0.08-0.09 in RMSE. Among them, the support vector machine regression model has the highest accuracy (prediction set R ² =0.91, RMSE=0.07; validation set R ² =0.89, RMSE=0.08), it will be as the best monitoring model of UAV for monitoring cotton disease severity.