利用Shannon熵参数的极化干涉SAR图像非监督分类

Unsupervised PolInSAR Image Classification with Shannon Entropy Parameters

本文利用Shannon熵参数,提出了一种极化干涉SAR图像非监督分类方法.Shannon熵是度量雷达照射媒质无序程度的物理量,它可表示成散射强度部分熵、极化部分熵以及干涉部分熵之和.本文利用Shannon熵分解出的干涉项参数结合Wishart最大似然聚类将SAR图像分成初始若干类,结合Shannon熵中散射强度项及极化散射项参数对初始类进行细分,再用层次聚类方法将这些类合并到需要的类数.运用Oberpfafenhoffen地区的PolInSAR数据进行分类,并与L.Ferro-Fmail提出的极化干涉分类方法的结果进行比较,实验结果证明了本文方法的有效性.

An unsupervised classification method making use of Shannon entropy parameter is proposed for polarimetric interferometric SAR（PolInSAR） data.Shannon Entropy measures the statistical disorder of the medium illuminated by the radar,and it can be decomposed into the the sum of three different terms that respectively represent the contribution of intensity,polarimetry and interferometry.Interferometric term parameter of the Shannon entropy is combined with Wishart maximum likelihood clustering to initially segment the SAR image into several classes.Those initial classes are further divided into several subclasses through the intensity term and polarimetric term parameter of the Shannon entropy,and then each initial classes are merged to desired number of classes through agglomerative hierarchical clustering method.