Feature fusion and recognition of potato disease images based on improved fractional differential mask and FPCA

For the problem of difficult location and recognition of potato diseases, we propose a method for potato leaves feature fusion and disease recognition which based on the improved fractional differential mask and fractional principal component analysis (FPCA). Firstly, the method preprocess potato leaf images by using improved fractional differential mask, and segment disease affected areas by adaptive threshold method. Secondly, fuse features from affected areas like color, shape and texture by fractional principal component analysis (FPCA). Finally, recognize potato disease images by support vector machine (SVM). We conducted recognition experiments on potato leaf images from those are affected by early blight or late blight, the results show that improved fractional differential mask and FPCA can effectively improve the recognition rate of potato disease images. Therefore, this paper use improved fractional differential mask, FPCA and SVM to recognize potato disease images, the recognition accuracy reached 98%.