Prediction of tomato plants infected by fungal pathogens at different disease severities using E-nose and GC–MS

Disease is a serious problem in tomato plants, causing huge economic losses. Disease detection, as the premise of protection, is important. This paper employed Electronic nose (E-nose) and Gas Chromatography–Mass Spectrometer (GC–MS), as an auxiliary technique, to predict disease type and its severity in the tomato plant. Twenty-five volatile constituents were identified using GC–MS. Their concentrations were calculated and showed the difference in different groups. Furthermore, the results of E-nose and GC–MS were compared and showed a good correlation. Moreover, the possibility of E-nose in classifying tomato plants infected with different types and severities of disease either respectively or together was proved based on either Principal Component Analysis (PCA) or Discriminant Functions Analysis (DFA). Then, Backpropagation neural network (BPNN) was introduced and showed that the correct classification rates were 98.3% for the training set and 97.5% for the testing set for predicting disease type and severity. Moreover, 100% correct classification rate was obtained for the diseased groups, which was very meaningful for the prevention of disease spread and met actual application needs. This study demonstrates the feasibility of E-nose in predicting tomato plants infected with disease in different types and severities.