EARLY DIAGNOSIS OF BOTRYTIS CINEREA INFESTATION OF TOMATO PLANT BY ELECTRONIC NOSE

Y. Sun,J. Wang,S. Cheng
DOI: https://doi.org/10.13031/aea.12748
2018-01-01
Applied Engineering in Agriculture
Abstract:Early diagnosis of disease is important for loss control. It is much easier to manage and prevent disease from spreading in this period. This study employed electronic nose (E-nose) for early diagnosis of Botrytis cinerea infestation of tomato plant. Gas Chromatography-Mass Spectrometer (GC-MS) was applied for proving the potential of E-nose detection and taken as the evidence for determining the range of parameters of Kernel Principal Component Analysis (KPCA). Then, the way to seek the best parameter (the type of kernel, kernel parameter, and the number of principal component) of KPCA for the prediction of time of tomato plant under disease attack was introduced. The results showed that when the type of kernel was Polynomial, the kernel parameter was 2, and the number of principal component was 17, the highest correct discrimination rate of Linear Discriminant Analysis (LDA) was obtained, which was as high as 100%. Furthermore, multiple linear regressions (MLR) was employed and the results showed that MLR combined with KPCA obtained excellent performance. This study demonstrated that it was feasible for early diagnosis of Botrytis cinerea infestation of tomato plant by E-nose and the model for predicting the infestation time was presented.
What problem does this paper attempt to address?