Early Detection and Continuous Quantization of Plant Disease Using Template Matching and Support Vector Machine Algorithms

Abstract

This paper present a novel method for robust and early Cercospora leaf spot detection in sugar beet using hybrid algorithms of template matching and support vector machine. We adopt three-stage framework to achieve our research target: first, a plant segmentation index of G-R is introduced to distinguish leaf parts from soil-contained background for automatic selection of initial sub templates, second, we adopt a robust template matching method called orientation code matching (OCM), which could not only realize the continuous and site-specific observation of disease development, but also shows its excellent robustness for nonrigid plant object searching in scene illumination, foliar translation and small rotation, afterward, we employ a machine learning method of support vector machine (SVM) for robust and early disease classification by a color-based feature named two dimensional (2D) xy-color histogram, which has stable ability to classify disease against various illumination changes. The indoor experiment results demonstrate the robust performance of our proposed method for early disease detection against complex changes of external environment.