Multi-model Approach for Grading & Classification of Paddy Leaf Diseases

Abstract

Paddy is one of the major crops in world. The yield of the crop is directly related to type of disease affected and level of infection. Detecting and classifying diseases automatically or semi-automatically will help the farmers to take necessary steps and to improve the crop yield.  In this paper, we are proposing a system to classify and grade the level of diseases affected to paddy leaves. The proposed system consists of four major steps. Initially in the first step, the system extracts leaf part from input image by eliminating background in L*a*b color space. In second step, proposed an algorithm to segment and binarize the input image in YCbCr color space, this algorithm extract the disease part and convert to binary image for further process. In third step, extracted features like mass, margin, shape, texture and color features based on the paddy leaf disease characteristics. These extracted features are fed to support Vector Machine (SVM) classifier to classify the paddy leaf diseases like, rice blast, brown spot, leaf blight, hispa and healthy leaves. In step four, Fuzzy Inference System (FIS) is used to grade the level of disease by considering infection percentage and maximum radius of disease part by setting 9×6 fuzzy rules. Finally, the performance of proposed segmentation & binarization algorithm is evaluated by extracting GLCM features and also by extracting proposed features and achieved 95.12% & 96.54% efficiency respectively. The similarity percentage and disagreements between manual prediction and Fuzzy Inference System to grade the level of disease affected is 90.66%, and 9.34% respectively.