Abstract:

In view of a lot of mixed image pixels contained in remote sensing images classification, fuzzy clustering support vector machine （SVM） was introduced to deal with the remote sensing images unmixing. In the proposed technique, GustafsonKessel is used to select the useful sample points for improving
 the classification performance of support vector machine. The effectiveness of the proposed method was evaluated through the forest cover remote sensing classification. The experiment shows that the accuracy of mixed pixels classification can be increased by applying the learning scheme, compared with that of traditional SVM classification method.

Key words: mixed pixels, GustafsonKessel fuzzy clustering, support vector machines （SVM）, remote sensing classification