Coastal Land Covers Classification of High-Resolution Images Based on Dempster-Shafer Evidence Theory

Integration of spectrum, texture and shape information, evidence theory is introduced to land covers classification of high-resolution images, and an object-oriented land covers classification method of high-resolution images based on Dempster-Shafer evidence theory is proposed. Firstly, for image objects, four kinds of indexes are selected as attributes to discriminate different land cover types, which are shape index, normalized difference vegetation index, normalized difference water index and entropy, respectively. Secondly, from the attributes input, belief functions of all the land cover types are calculated, and then classification rules formation as ldquoattributes-> categoryrdquo are extracted by maximizing belief value. Lastly, according to the rules mined, automatic classification of land covers can be realized.