An Implementation of a Land Cover Pattern Classification System for Remotely Sensed Data by Using Neuro-Fuzzy Algorithm

Abstract

In this paper, we implement a land cover pattern classification system for remotely sensed data by using a neuro-fuzzy algorithm, and compare it with the conventional methods of the Back-Propagation learning and the Maximum-Likelihood algorithm. The neuro-fuzzy pattern classifier has a 3-layer feed-forward architecture that is derived from a generic fuzzy perceptron. The digital image used in our research was acquired with the AMS (Airborne Multispectral Scanner). We determine the eight classes covered the majority of land cover feature on Daeduk Science Town. The results show that the proposed classifier is considerably more accurate to the mixed composition area with complex classes.