应用单类分类提取极化SAR影像中的倒塌建筑物

USING ONE-CLASS METHOD TO EXTRACT COLLAPSED BUILDINGS FROM POLARIMETRIC SAR IMAGES

通过引入单类分类方法,将倒塌建筑物作为唯一目标样本进行提取。介绍了两种基于支持向量机的单类分类方法,并选择最小超球体单类支持向量机用于震后单时相Pol SAR影像中的倒塌建筑物提取实验,结果表明单类分类方法能够融合多种特征快速提取倒塌建筑物,且能保证一定的提取精度,是一种行之有效的震害提取方法。

The rapid and accurate acquisition of collapse building can guide the effective implemen- tation of emergency rescue, and reduce the damage and casualties at the maximum extent. In this paper, the One-Class method is introduced to extract the collapsed buildings which are the only tar- get of our interest. Two kinds of the One-Class classification method based on support vector ma- chine （SVM） are described in this paper. The method of minimum hyperspherical One-Class sup- port vector machine is used for the experiments of collapsed buildings extraction only from only post- earthquake PolSAR imagery. The results showed that One-Class approach can fuse multiple features and rapidly extract collapsed buildings, at the same time the extraction accuracy is not very low, so the One-Class is a kind of effective method for earthquake collapsed buildings extraction.