Plant Diseased Leaf Identification and Classification using Combination of Rough k-means and Super-pixel with Discriminative Features

It is difficult for farmers to monitor their product and control all parameters manually. Image processing technique can be used for overcoming this problem. In this paper we proposed a new method for plant diseases detection based on image processing technique. Our proposed method is based on combining super-pixel and Rough k-means clustering. We use the hybrid clustering method in order to obtain segmented image of plant leaf. The Pyramid Histogram Oriented Gradient (PHOG) is extracted from segmented image with three level as a part of feature vector. In addition to PHOG, Gray Level Co-occurrence Matrix (GLCM) is created from the segmented image and useful statistics are extracted as texture features. Finally, we use C-SVM as classifier in order to perform final recognition. The proposed algorithm behavior is analyzed by simulation and compared with other state of arts in case of validating enhancements.