应用分水岭变换与支持向量机的极化SAR图像分类

Polarimetric SAR Image Classification Using Watershed-Transformation and Support Vector Machine

结合分水岭变换与支持向量机的特性,提出一种新的极化SAR图像分类算法。其基本思想是先通过分水岭变换及区域合并处理,将极化SAR图像分割成一系列同质区;再以同质区为基本单元,进行特征提取及样本选择后采用支持向量机分类。实验结果表明,该算法可有效降低相干斑对分类的影响,与传统基于像素的SVM算法相比,其分类精度有显著的提高,且结果也更易于理解。

Considering the properties of watershed-transformation and support vector machine,a method for classifying polarimetric SAR image is proposed in this paper.First,polarimetric SAR image is segmented into a series of homogenous regions through watershed transformation and region merging process.Then,region-based classification is performed by utilizing support vector machine after feature extraction and sample selection.Experimental results show that the proposed classification method depresses speckle effectively,when in comparison with traditional pixel-based SVM algorithm,the classification accuracy is improved by dramatically and more interpretable result can also be achieved.