Scattering feature decomposition and SVM classification of multi-angle polarimetric SAR images

In the understanding and interpretation of polarimetric synthetic aperture radar (PolSAR) images, terrain classification is one of the most important applications. The paper studies the terrain classification of multi-angle polarimetric SAR images based on non-analytic scattering models. The feature model decomposes multi-angle polarimetric SAR images by three decomposition methods to obtain characteristic parameters, and finally classifies. First, we quantify and rank media polarimetric scattering dissimilarity over all aspects. In addition, for the multi-polarization feature decomposition model, a general multi-angle polarization feature first-order difference sequence description method and coding method are proposed, including Yamaguchi four-component decomposition, Krogager decomposition and H/A/Alpha decomposition, which are decomposed to obtain multi-dimensional characteristic parameters. Finally, the feature sequences are classified by support vector machines(SVM). The 360-degree observation experiment of Pband polarimetric SAR is carried out to verify the effectiveness of the method and reveal the application potential in the terrain classification of features.