Abstract
A Spatial land use/cover classification algorithm for high resolution images such as TM and SPOT utilizing a fuzzy concept has been developed. This algorithm is composed of 2 stage classifications. In the first stage, conventional point wise classifiers are used. In this study, a supervised maximum likelihood classifier was used. Classified classes in the first stage are called land cover elements. In the second stage, each land use/cover category is thought to be a fuzzy set and is composed of land cover elements. Two kinds of fuzzy estimations, i. e., min-max inference and algebraic sum-max inference were tested. As a result, the algebraic sum-max algorithm showed about 5% better classification accuracy than a point wise classifier.
