Diagnostic Models For Tobacco Mosaic Virus And Potato Virus Y Of Tobacco Based On Spectral Analysis

Tobacco mosaic virus(TMV)and potato virus Y(PVY)are important viruses in tobacco,which cause serious losses in tobacco production.Accurate monitoring of diseases can provide scientific basis for disease control.In this study,spectral data of TMV and PVY-infected tobacco were obtained by ASD non-imaging spectrometer in greenhouse and field,and the measurement method of spectral data of tobacco leaf was optimized.The discrimination models of TMV and PVY classification and severity discrimination were established based on stepwise regression,support vector machine,K nearest neighbor classification and random forest algorithm.The main results are as follows.1.The measurement method of tobacco leaf spectral data was optimized.According to the spectral characteristics of different parts of tobacco leaf,the data of different parts of tobacco leaf measured at different frequencies were obtained by spectral techniques.It was found that the optimum measuring times to reflect the characteristics of leaf spectrum data were 3 times(18 sampling points),which laid a method foundation for establishing the discriminant model of TMV and PVY disease types and severity.2.The discrimination models of TMV and PVY diseases in tobacco were established.In the visible light band(400 ～ 700 nm),the leaf spectral reflectance of PVY-infected plants was higher than that of TMV-infected plants and healthy plants.In the near-infrared band(700 ～ 1300 nm),the spectral reflectance of leaves of PVY-infected plants was lower than that of healthy plants.The spectral reflectance of TMV-infected plants was higher than that of healthy tobacco plants and PVY-infected plants.The sensitive bands of TMV and PVY disease classification were selected by ranking-based clustering method(E-FDPC),and the discrimination models of TMV and PVY disease were established based on support vector machine algorithm.The overall accuracy of the models was 85.71%.3.The discrimination models of TMV and PVY disease severity were established.With the increase of TMV and PVY disease severity,the leaf spectral reflectance increased gradually in visible light band.In near infrared band(700 ～ 1300 nm),with the increase of disease severity,the leaf spectral reflectance decreased for PVY-infected plants and changed irregularly for TMV-infected plants.The sensitive wavelengths were 453,518,540,572,708,715,775,780 and 1000 nm,which were significantly correlated with the severity of for PVY-infected plants.Based on the ranking clustering method,600 sensitive bands for PVY and TMV disease classification were selected.Based on the stepwise regression method,the discrimination model of PVY disease severity was established.The overall accuracy of the model was 83.84%.Based on the support vector machine algorithm,K nearest neighbor classification algorithm and random forest algorithm,the discrimination models of PVY and TMV disease severity were established.The overall accuracy of the models was 90.72%,86.75% and87.78%,respectively,the support vector machine model was the optimal model.To sum up,the data of tobacco leaf were obtained by spectral instrument,and the measurement method of the spectral data of tobacco leaf was optimized.The classification and severity models of TMV and PVY diseases were established respectively,and the optimal algorithm was support vector machine algorithm.This study can provide a basis for monitoring,early warning and integrated management of tobacco virus diseases.