基于Krogager分解和SVM的极化SAR图像分类

Classification of Polarimetric SAR Image Based on Krogager Decomposition and SVM

目标分解包括基于Sinclair矩阵的相干目标分解和基于Mueller矩阵的部分相干目标分解,Krogager分解即属于相干目标分解,它可以将任一对称Sinclair矩阵分解为球散射体、二面角散射体和螺旋体3个分量,这是极化合成孔径雷达(Synthetic Aperture Radar,SAR)图像特征提取的有效途径。把3个分量的分解系数作为极化散射特征,由其组成样本向量,运用基于统计学习理论的支持向量机(Support Vector Machines,SVM)设计多类分类器,提出了一种极化SAR图像分类算法,并对实测极化SAR数据进行分类实验。结果表明,将Krogager分解和SVM分类器结合起来,对极化SAR图像进行分类是可行和有效的,并且选择不同的参数得到的分类结果差别很大,验证了参数选择在SVM分类器中的重要作用。

Target decomposition comprises coherent target decomposition based on Sinclair matrix and partcoherent target decomposition based on Mueller matrix, Krogager decomposition is a kind of coherent target decomposition. By Krogager decomposition, a symmetric Sinclair matrix can be decomposed into three components： sphere component, diplane component, and helix component. This is an important way to extract properties from polarimetric synthetic aperture radar （SAR） image. In this paper, decomposition coefficients of three components are acted as polarimetric scattering properties. Samples are composed of three decomposition coefficients. Multi-class classifier can be designed using support vector machines （SVM） based on statistical learning theory, a new algorithm of target classification is proposed, and classification experiments to polarimetric SAR data are done. The results indicate it is feasible and efficient to classify polarimetric SAR image by combining Krogager decomposition and SVM. Discrimination of classification results by choosing different parameters is rather big. Thus, parameters selecting is very important to SVM classifier.