Superpixel based roughness measure for cotton leaf diseases detection and classification

Color image segmentation is very important for separating an object of interest from given input image. For cotton leaf disease detection, an infected part of leaf must be separated out for further classification. This paper proposed a technique for cotton leaf diseases detection and classification using the concept of roughness measure and simple linear iterative clustering. An optimum number of superpixel group are formed using roughness measure for extracting region of interest of cotton leaf. Gray level co-occurrence matrix features are extracted from detected region. Support vector machine, a supervised machine learning algorithm is used to classify cotton leaf into four different categories as Alternaria diseases, Bacterial diseases, White flies, and Healthy cotton leaf. Proposed algorithms demonstrated the average classification accuracy of 94% with the available database.