Early Detection and Classification of Tobacco Leaves Inoculated with Tobacco Mosaic Virus Based on Hyperspectral Imaging Technique

Abstract

Abstract. Tobacco mosaic virus (TMV) is a severe worldwide disease and capable of greatly reducing tobacco quality and yield. It is vital to identify the tobacco disease inoculated with TMV at the early stage of infection for preventing and managing plant diseases. The main contribution of this paper is a procedure for the early detection and classification of tobacco disease with different machine-learning algorithms based on hyperspectral imaging technique. Images from healthy leaves and leaves inoculated with TMV for a period of 7 days after inoculation were acquired per two days by a hyperspectral imaging system with the wavelength region of 380-1,023 nm. Spectral reflectance of pre-defined region of interest (ROI) from hyperspectral images was extracted by ENVI software. Successive projections algorithm (SPA) were used and evaluated for effective wavelengths (EWs) selection. In addition, different machine-learning algorithms i.e. support vector machine (SVM), back propagation neural network (BPNN), extreme learning machine (ELM), least squares support vector machine (LS-SVM), partial least squares-discrimination analysis (PLS-DA), linear discriminant analysis (LDA) and random forest (RF) were developed to quantitatively detect and classify disease stages using EWs. Most classification models showed acceptable results, with the identification rate was over 85%. The discrimination between healthy tobacco leaves and diseased leaves (healthy, 2 days post infection (DPI), 4 DPI, 6 DPI) resulted in classification accuracies up to 95% with BPNN and ELM models. Hence, the overall results clearly demonstrated that hyperspectral imaging has the potential as a fast and non-invasive method with machine-learning algorithms for presymptomatic detection of the plant diseases.