Use of fuzzy logic for unsupervised classification of remotely sensed images

Fuzzy logic offers an opportunity to overcome some of the limitations attached to conventional classifications by providing weighted description, by means of membership values, of each pixel to generic categories of land occupation areas. These membership values are obtained by adjustment of pre-defined membership functions, expressed by means of linguistic predicates, to experimental conditions using solely information contained in the image itself. The image is then classified according to the membership values using different fuzzy rules. A winter time TM image of Brittany (France) was processed according to this method.