结合Singh四分量分解的高分三号全极化SAR影像分类

不同于一般分类算法基于像素统计的分类，忽略了地物的散射特性，文中提出了一种保持地物散射特性的分类方法。这种方法将Singh提出的Singh四分量分解与基于复Wishart分布的最大似然分类器相结合，对高分三号全极化影像进行分类。利用Singh四分量分解获得表面散射、体散射、二次散射和螺旋体散射，然后将前3种基础散射分别划分为多个聚类，根据复Wishart距离进行类间合并，直到获得指定类别数，输入复Wishart分类器进行迭代分类，最后进行类别合并获得最终分类结果。试验表明本文算法具有较好的分类效果且验证了利用高分三号全极化卫星数据进行影像分类的可行性。

Classification of GF-3 fully polarimetric SAR image combined with the Singh four-component decomposition

Different from general classification algorithm, pixel statistics based classification ignores the scattering characteristics of objects. In this paper, an classification method for preserving the scattering characteristics of objects was proposed. This method combined the Singh four-component decomposition proposed by Singh with the maximum likelihood classifier based on the complex Wishart distribution for classification of GF-3 polarimetric images. Surface scattering, volume scattering, double-bounce scattering and helix scattering were obtained by Singh four-component decomposition, then the first three basic scatterings were divided into multiple clusters respectively, and the inter-class merging was performed according to the complex Wishart distance until the specified number of categories was obtained. The Wishart classifier was used foriterative classification and merged the categories at last to obtain the final classification result. Through experiments, it was proved that the algorithm proposed in this paper had a good classification performance and verified the feasibility of using GF-3 satellite data for image classification.