The disease severity estimation of bacterial spot disease of processing tomato based on hyperspectral remote sensing

This paper will determine the canopy spectral reflectance and disease severity of processing lomato bacterial spot disease in field and disease nursery. The canopy spectral reflectance In different varieties and different growth stages are transformed into first-order differential, second-order differential, and inverse logarithm spectral reflectance. We analyze the correlation of the disease severity and primitive spectral reflectance, first-order differential, second-order differential, and inverse logarithm spectral reflectance, in order to look for the sensitive band of diseased processing tomato. The sensitive spectrum is established for the model of disease severity and test. The result shows: according to the R² and F value of regression between disease severity and primitive spectral reflectance, first-order differential, second-order differential, and inverse logarithm spectral reflectance, R_758nm, R_773nm, R_779nm, R_784nm, R_2493nm, FD_379nm, FD_755nm,FD_868nm, FD_874nm, FD_877nm, SD_456nm, SD_521nm, SD_578nm, SD_617nm, SD_778nm, 1/log_(2487nm), 1/log_(2488nm), 1/log_(2493nm), 1/log_(2494nm), 1/log_(2497nm) are sensitive spectrums. The partial least-squares regression model of second-order sensitive spectral reflectance and disease severity is the best model, because there are explain ability of disease severity of 0.824 and RMSE was 0.127.The R² of fitting with observation value and predictive value is 0.723. It had practical value to monitor the large area of processing tomato bacterial spot disease using hyperspectral.