Remote sensing images classification using fuzzy-rough neural network

In remote sensing images classification, the boundaries between different classes are vague and it is often difficult or impossible to acquire all of the necessary essential features for precisely classification. So both the fuzzy uncertainty and rough uncertainty are presented. Based on fuzzy-rough set theory, a fuzzy-rough neural network (FRNN) is designed for remote sensing images classification. In the FRNN classification algorithm, fuzzy set, rough set and neural network technique are combined. Fuzzy-rough function is used as membership function of the FRNN and integrates the ability of processing fuzzy and rough uncertainty information, which endue the FRNN classifier with better capability of learning and self-adapt. Experimental results show that the proposed classification algorithm can be used in remote sensing images classification, and its classification precision is superior to that of the conventional maximum likelihood algorithm and radial basis function neural network (RBFNN) algorithm.