Inconsistent Cluster Analysis With Disease Feature Enhancement (ICADFE) For American Cotton Leaf Disease Recognition

The broad leaves of cotton plant carry various visible disease symptoms. The ability of visual analysis by experts motivated the development of the plant disease recognition model. There are several visual feature descriptors, which can be primarily distinguished on the basis of pattern, texture or color. This system has been developed for the convenience of the farmers, who can avail the benefit by submitting the pictures of infected cotton leaves on the interface and the plant disease recognition system will return type of disease. In this paper, a dynamic feature descriptor is designed with inconsistent cluster analysis (ICA) and disease feature enhancement (DFE), which are combined as hybrid descriptor known as ICADFE for the recognition of the cotton plant disease. The ICADFE is found to improve the detection accuracy (approx. 80%), precision (approx 95%) and f1-measure (approx. 88%) on average in comparison with traditional shape and texture based feature descriptors such as scale invariant feature transform (SIFT), speeded up robust features (SURF) and fast retina keypoints (FREAK) with multicategory SVM (mSVM) for disease recognition