Classification of earth terrain in polarimetric SAR images using neural nets modelization

摘要

Two supervised classification procedures are presented and applied to relative polarimetric SAR images in order to identify the different Earth terrain components. The first one is the classical Bayes classifier, and the second is an original polarimetric method based on a neural network modelization. The subject of this paper is to show that it is possible to classify polarimetric data by using neural network techniques, without knowing the a-priori statistic distributions of the different classes. The purpose being to show that POLARIMETRY and IA theories can become complementary sciences.