PolSAR Sea Ice Classification Application Base On Feature Fusion

Polarimetric Synthetic Aperture Radar (PolSAR), has the property of day and all-weather. It is one of the most important remote sensors for sea ice classification at present. PolSAR can get the polarization scattering properties of targets on the basis of achieving the intensity of target. It can provide data base to improve the classification accuracy of sea ice. It is significant to object extraction. Because of the formation of sea ice is very complex, the difference between different types of ice varies less than land target, so only use polarization feature to classify is easy to cause the misclassification of sea ice. Separately using polarization and texture feature exist two serious problems:First, it may lose classification because of environmental factors and electromagnetic damping. Second, the calculation of gray level co-occurrence matrix may cost a lot of time.The polarization feature focuses on the backscatter coefficient of the current pixel, and the texture feature is mainly based on adjacent pixel points in spatial distribution. Aiming at the complexity of sea ice classification, this paper focus on the study of the features fusion of sea ice classification.First of all, analyze the texture and polarization features of sea ice. Then classify the sea ice by using the analyzed characteristics based on feature fusion of multiband supervised classification model. In this paper three scene PolSAR images on C-band and L-band taken by SIR-C are analyzed. And the experimental results verify the validity of this method. Sum up, this article main work and achievements are as follows:(1) In order to evaluate the effectiveness of the various characteristics, in this paper polarization and texture feature are analyzed and screened. By analyzing the three scene SIR-C sea ice data, select a set of features suitable for sea ice classification.(2) MSCM has been proposed.(3) By analyzing three scene PolSAR images on C-band and L-band taken by SIR-C. It is proved that using fusion feature to classify sea ice can effectively improve the classification precision of sea ice, and get effective classification.