Comparison with two classification algorithms of remote sensing image based on neural network

Abstract

The traditional approaches of classification are always unfavorable in the description of information distribution. This paper describes the BP neural network approach and the Kohonen neural network approach to the classification of remote sensing images. Two algorithms have their own traits and can be good used in the classification. A qualitative comparison demonstrates that both original images and the classified maps are visually well matched. A further quantitative analysis indicates that the accuracy of BP algorithm is better than the result of the Kohonen neural network © Springer-Verlag 2004.