基于最优极化相干系数的倾斜建筑物解译研究

摘要

基于Freeman_Durden分解的全极化SAR影像分类方法能够较好地保持地物极化散射特性，但在分类的过程中，不能改变初始散射机制，导致分解结果对分类精度影响很大。在Freeman_Durden分解中，排列方向相对雷达飞行方向不平行的建筑物（简称为倾斜建筑物）常被分为体散射类型，使得该类建筑物往往被误分为植被。通过分析建筑物在SAR影像中的后向散射特性，利用建筑物具有较高相干性的特点，引入最优极化相干系数，在目标分解的基础上通过阈值分割将两者区分开来，进而提高反射非对称性人工目标的分类效果。通过使用E—SAR系统在德国DLR附近Oberp—faffenhofen地区获取的L波段PollnSAR影像和国内X—SAR系统在海南陵水地区获取的X波段PolInSAR影像进行试验，证明该方法能够有效地将与雷达飞行方向不平行的建筑物与森林区分开。%It could preserve polarimetric scattering characteristics for classification of polarimetric synthetic aperture radar image based on Freeman_Durden decomposition, but the classification accuracy is susceptible to target decomposition due to that the scattering mechanisms could not be changed in the process of classification. Buildings aligned not along with the flight direction of radar belong to volume scattering in Freeman_Durden decomposition. And it was difficult to distinguish oblique buildings from vegetation. According to the high coherence characteristic of building, optimal polarimetric coherence coefficient was introduced to the new algorithm by analyzing the backscattering characteristics of building, to improve the classification accuracy of tilted building and forest on the basis of target decomposition. The experiment results indicate the effectiveness of the algorithm by using the L-band PolInSAR images of Oberpfaffenhofen around DLR of E-SAR and X band PolInSAR images of Lingshui in Hainan province of domestic X-SAR.