NEURAL-FUZZY CLASSIFICATION FOR SEGMENTATION OF REMOTELY-SENSED IMAGES

An unsupervised classification technique conceptualized in terms of ne ural and fuzzy disciplines for the segmentation of remotely sensed ima ges is presented, The process consists of three major steps: 1) patter n transformation; 2) neural classification; 3) fuzzy grouping. In the first step, the multispectral patterns of image pixels are transformed into what we call coarse patterns, In the second step, a delicate cla ssification of pixels is attained by applying an ART neural classifier to the transformed pixel patterns. Since the resultant clusters of pi xels are usually too keen to be of practical significance, in the thir d step, a fuzzy clustering algorithm is invoked to integrate pixel clu sters. A function for measuring clustering validity is defined with wh ich the optimal number of classes can be automatically determined by t he clustering algorithm, The proposed technique is applied to both syn thetic and real images. High classification rates have been achieved f ar synthetic images. We also feel comfortable with the results of the real images because their spectral variances are even smaller than tho se of synthetic ones examined.