Sea ice classification using dual polarization SAR data

Abstract

Sea ice is an indicator of climate change and also a threat to the navigation security of ships. Polarimetric SAR images are useful in the sea ice detection and classification. In this paper, backscattering coefficients and texture features derived from dual polarization SAR images are used for sea ice classification. Firstly, the HH image is recalculated based on the angular dependences of sea ice types. Then the effective gray level co-occurrence matrix (GLCM) texture features are selected for the support vector machine (SVM) classification. In the end, because sea ice concentration can provide a better separation of pancake ice from old ice, it is used to improve the SVM result. This method provides a good classification result, compared with the sea ice chart from CIS.