基于散射成分一致性参数的极化SAR图像分类

Scattering component consistency based parameter for polarimetric SAR image classification

散射熵能较好地反映目标散射的随机性,但忽略了相干矩阵特征分解后3个相干散射成分之间的关系。为了更充分地利用极化信息提取更有效的特征,该文提出一种描述目标散射成分一致性的新参数,并利用该参数进行图像分类。新参数融合了相干矩阵的特征值分布信息与各正交散射成分之间的相似性信息,反映了目标的整体散射机制接近于某种单一相干散射的程度。利用该新特征替代散射熵,先对AIRSAR的旧金山L波段数据进行初始分割,然后进行基于Wishart分类器的迭代调整。实验结果表明:利用该特征能够更准确地实现图像分类,展现地物细节,从而证实了该特征的有效性。

The scattering entropy accurately describes the randomness of a scattering medium, but analyses do not use the relationships between the three eigenvectors representing the different coherent scatterings. More polarization information is extracted by a parameter describing the consistency of the scattering components to classify of polarimetric SAR images. The parameter contains information on the eigenvalue distributions and similarities between the coherent scattering components and represents the closeness of the scattering to simplex coherent scattering. The AIRSAR I-band polarimetric image of San Francisco is segmented using this parameter instead of the scattering entropy and then adjusted by a Wishart classifier. Tests demonstrate the effectiveness of this parameter to improve the classification and object details.