基于非负特征值分解和SVM的极化SAR图像分类

摘要

The nonnegative eigenvalue constraint has the advantage on determining whether a polarization matrix can characterize the scattering mechanism of the target. If a matrix cannot satisfy the nonnegative eigenvalue constraint, the scattering mechanism of the target won't be characterized and the polarization information of the ground object scattering won't be obtained. The polarization characteristics of the de-composition which ignores the constraint in the polarization decomposition are meaningless. This paper extracted the plane scattering, even scattering, volume scattering component by the nonnegative eigenval-ue decomposition method, proposed a method of SAR image classification based on SVM classifier, and carried out the classification experiments on AIRSAR_SanFrancisco data, the classification results of which was compared with that of H/A/ alpha－Wishart classification. The results showed that the classifi-cation method of nonnegative eigenvalue decomposition combined with SVM classifier was feasible, effec-tive and good classification accuracy.%非负特征值约束的优势在于它可以判定一个极化矩阵能否对目标的散射机制进行表征.若一个矩阵不能满足非负特征值约束条件,则不能表征目标的散射机制,也就没有地物目标散射的极化信息.在非负特征值约束理论之前的极化分解中忽略了这个条件,导致分解的极化特征没有意义.根据满足非负特征值约束的非负特征值分解方法,提取出平面散射、偶次散射、体散射分量,结合支持向量机分类器,提出了一种SAR图像分类方法,并以AIRSAR_SanFrancisco数据进行分类实验,且将分类结果与H/A/α－Wishart分类结果进行对比分析.结果表明,基于非负特征值分解和支持向量机分类器结合的分类方法可行有效,且具有很好的分类精度.