A Hierarchical Fuzzy Classification Scheme For Remote Sensing Data

Abstract Land covers mix and high input dimension are two important issues that affect the classification accuracy of remote sensing images. Fuzzy classification has been developed to represent the mixture of land covers. Two fuzzy classifiers of fuzzy rules-based (FRB) and fuzzy neural network (F’NN) were studied to illustrate the interpretability of fuzzy classification. Based on the FNN classifier, a hierarchical FNN (HFNN) classifier was proposed to solve the problem of the high input dimension. It was compared with the FNN classifier on the land cover classification of a Landsat 7 ETM+ image over Rio Rancho, New Mexico and was proved to be the best combination of better classification accuracy with shorter CPU time requirement.