Fusion and classification of remote sensing images using fuzzy logic

Summary form only given. We propose a method for fusion and classification of remote sensing images by using fuzzy logic. The purpose is to overcome the problems that are often encountered in practice. In fact, when statistical classifiers are used for classifying remote sensing images, it is often implicitly assumed that the training samples used to train the classifier represent the true classes. This assumption may not be valid in practice for several reasons. In addition, the statistical classifiers cannot handle the uncertain and imprecise aspects relative to the remote sensing images since they have not the mechanism of modelling and processing the uncertainty. The basic idea is to use the ability of the fuzzy logic to model and process the uncertain and imprecise aspects relative to the remote sensing data and to perform well when there are not enough training samples that represent the true classes to accurately estimate the parameters of the classifier. The proposed method is tested and evaluated by using real data. It outperforms conventional statistical classifiers and it may be used for various applications with multisource data.