全极化SAR极化特征谱应用研究

在分析特征值分解结果,全部散射机制组合和极化特征谱性质的基础上,提出基于3个特征谱参数的假彩色合成方法,可以更加有效直观地反映地物散射特征,再对散射熵、散射角、反熵和4个极化特征谱参数进行特征选择分析,给出最佳的多维特征向量选择方案,从而实现传统遥感图像分类器如同ISODATA算法对极化SAR图像的分类。实验选择了一景Radarsat\|2标准全极化SAR数据,包含典型的城市、植被和水体三大类地物,实验结果表明:极化特征谱假彩色合成充分反映了各地物散射特征,特征谱和散射角组成了最佳特征向量,非监督分类结果表明:该方法克服了城市与植被在H\|Alpha平面上分布界限模糊的问题,分类精度高于H\|Alpha平面非监督分类,与Wishart-H-Alpha-A分类方法相当。

The Application Research of Polarization Characteristic Spectrum for Polarimetric SAR

This paper analyzes the eigenvalue decomposition,all combinations of scattering mechanisms and polarization characteristic spectrum,and then proposes a method of false color composition based on three kinds of characteristic spectral parameters,which can be more effective directly reflect the scattering feature.Then,the scattering entropy,scattering angle,anti-entropy and four parameters of polarization characteristic spectrum area are studied in feature space.And this paper gives the best options for multi-dimensional feature vector,in order to achieve the traditional classification algorithm to process polarimetric SAR images,such as ISODATA.This study selects a scene of Radarsat-2 polarimetric SAR data for test,including typical urban,vegetation and water.The experimental results show that the false color composition with polarization characteristic spectrum reflects the local feature of the scattering material.The unsupervised classification with three of characteristic spectrum and scattering angle show that the proposed method overcome the defects in H-Alpha plane,which causes blurring segment between city and the vegetation.Classification accuracy is higher than the unsupervised classification with H-Alpha plane,but has the similar effect with Wishart-H-Alpha-A classification algorithm.