FUZZY SUPERVISED CLASSIFICATION OF REMOTE-SENSING IMAGES

In the conventional remote sensing supervised classification, training information and classification results are represented in a one-pixel-one-class method. Class mixture cannot be taken into consideration in training a classifier and in determining pixels' membership. The expressive limitation has reduced the classification accuracy level and led to the poor extraction of information. This paper describes a fuzzy supervised classification method in which geographical information is represented as fuzzy sets. The algorithm consists of two major steps: The estimate of fuzzy parameters from fuzzy training data, and a fuzzy partition of spectral space. Partial membership of pixels allows component cover classes of mixed pixels to be identified and more accurate statistical parameters to be generated, and, in turn, a higher classification accuracy to be achieved. Results of classifying a Landsat MSS image are presented and their accuracy is analyzed. © 1990 IEEE