STUDY ON PLANT DISEASE IDENTIFICATION TECHNOLOGY BASED ON REFLECTANCE SPECTRA

It is desirable to develop a technology that enables anyone to diagnose the plant diseases properly and easily to minimize the crop damage from the diseases. In this study, we aimed to establish a simple plant disease identification technology based on plant reflectance spectra measured with a spectrometer and attempted to classify healthy cucumber and mosaic disease caused by zucchini yellow mosaic virus (ZYMV) at the early and late stages of infection. Spectral data were measured with a Spectrum Catcher (Polar Star Space, Japan), and the reflectance spectral patterns of healthy leaves and ZYMV infected leaves were discriminable significantly by statistical analysis. Therefore, the classification learning was performed based on the selected vegetation indexes calculated from the reflectance and the experimental results showed a high classification accuracy of 91.4% in the early stage and 95.6% in the late stage. Furthermore, we examined and proposed a new normalized difference spectral index (NDSI) using two wavelengths, 656 nm and 658.2 nm, as an index that could discriminate diseases and the classification learning based on the NDSI achieved 77.2% accuracy that was better than that of the system using the conventional NDVI