Research on Yellow River Ice Supervised Classification Method Based on Polarimetric SAR Data

Abstract

According to the different contents of sediment and bubbles in the Yellow River ice, the Yellow River ice can be classified into three types: thermal ice, frazil ice and consolidated ice. Different impurity contents correspond to different microwave electromagnetic scattering characteristics. On the basis of measuring the ice thickness of the Yellow River in situ, this paper uses the Radarsat-2 full polarization SAR satellite data of Canada to do supervised classification of the Yellow River ice. Firstly, three classical minimum distance classifier, fisher classifier and Wishart classifier are used to classify the Yellow River ice. Then the classification effects of the three classifications are analyzed by using the confusion matrix. Finally, the advantages and disadvantages of the three classifications are compared and analyzed. The experimental results show that Wishart classifier has a good classification effect on the Yellow River ice, and the overall classification accuracy reaches 95%. The classification results provide input conditions for subsequent modeling of the Yellow River ice electromagnetic scattering and ice thickness retrieval.