RADARSAT-2全极化SAR数据地表覆盖分类

Land cover classification using RADARSAT-2 full polarmetric SAR data

全极化合成孔径雷达(SAR)能够测量每一观测目标的全散射矩阵,但地物分布的复杂性往往造成不同地物具有相似的后向散射信号特征,因而增加了地物信息提取的难度。文中基于北京地区的RADARSAT-2全极化雷达数据,在图像处理的特征分解的基础上,利用PolSARPro软件提取包含地物散射机理信息的各种极化参数,按H-α、A-α、H-A对全极化SAR影像进行基于散射机理的分类,继而将分类结果作为Wishart H/A/α、Wishart H/α的初始类别划分。最后,采用决策树分类算法对基于Wishart分布的监督分类及以上两种分类算法进行融合处理,从而实现地物的分类,并将分类结果与经典的分类算法进行对比分析,验证了文中方法的有效性。

Since polarimetric synthetic aperture radar (SAR) is capable of measuring each observation target full scattering matrix ,but because of the complexity of the feature distribution it is often caused by scattering signals of different objects with similar characteristics to the rear ,thus increasing the difficulty of feature information extraction .Based on RADARSAT-2 fully polarimetric radar data in Beijing ,on the basis of the characteristics of the image processing decomposition ,PolSARPro software is utilized to extract contains information about terrain scattering mechanism of polarization parameters ,according to H-α,A-α, H-A on full polarization SAR images can be classified based on the scattering mechanism .Then the results will be classified as Wishart H/A/αWishart H/αinitial category .Finally ,the decision tree classification algorithm is used based on supervised classification of Wishart distribution and the above two kinds of classification algorithms are put into fusion processing ,so as to achieve the classification of surface features .The classification results with the traditional classification algorithms are analyzed to prove the effectiveness of the proposed method .