文章基本信息

标题：Fuzzy Classification of Satellite Imagery by Neural Networks
本地全文：下载
作者：Eihan Shimizu
期刊名称：ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
印刷版ISSN：2194-9042
电子版ISSN：2194-9050
出版年度：1992
卷号：XXIX Part B7
页码：964-968
出版社：Copernicus Publications
摘要：The conventional method for classification of satellite imagery is based on Bayes' theorem. The applied condition is to be "eachpixel must belong to either of the classes". In other words, none of the pixels can belong to more than one class ('mixed' pixel), norbelong to none of the classes ('unknown' pixel). However, the 'mixed' pixel is necessarily existent in the case of satellite imagery.Also, the existence of 'unknown' pixel is inevitable as the number of class settings is restricted. This paper discusses the fuzzyclassification of the satellite imagery. The classes are defined as fuzzy sets in spectral space. With this the 'mixed' and 'unknown'pixels can be considered by the fuzzy set operations. It is difficult to directly give a membership function of the class fuzzy set in amulti-spectral space. Therefore, it is approximately estimated from the training data. By defining the membership function on the leastsquares criteria from the training data, I/O system equivalent to this function can be realized with back propagation algorithm of theneural network. The performance of the fuzzy classification is evaluated in comparison with the conventional supervised classification.The fuzzy classification method is able to provide a land cover classification superior to that derived from the conventional method.This paper also describes a method to effectively visualize the fuzzy classification result using ROB color composite
关键词：Fuzzy classification; Neural network; Back Propagation Algorithm; Visualization of Fuzzy Event