结合形态学属性断面与支持向量机的合成孔径雷达图像变化检测

Change detection of SAR images using morphologic attribute profile and support vector machine

针对传统合成孔径雷达(SAR)图像变化检测方法存在误差大、检测率低等问题,提出了一种基于形态学属性断面(MAP)的SAR图像变化检测方法。该方法利用MAP算法提取差异图像的几何结构特征,构造深入描述图像结构化信息的特征向量空间;在利用阈值法对图像进行分割的基础上,引入偏移因子,实现训练样本的自动选取;最后,用支持向量机(SVM)在多维特征空间中对图像进行变化类与非变化类的分类。实验结果显示:本文算法的检测结果优于基于高斯模型的KI阈值法(GM_KI)、基于广义高斯模型的KI阈值法(GGM_KI)和大津法(Otsu)等3种阈值法的检测结果,Kappa系数保持在0.87以上;当峰值信噪比(PSNR)介于[29,44]dB时,抗噪性能指标保持在0.97以上。这些结果证明了文中方法的有效性和优越性。

As classical change detection methods for Synthetic Aperture Radar （SAR） images have high error rates and low detection rates, a novel change detection method of SAR images based on Morphology Attribute Profile （MAP） was proposed. The MAP algorithm was employed to extract the geometric features of the difference images and a feature vector space was constructed to describe the image inherent structure. Then, the offsets were introduced to select the training samples automati-cally based on the segmentation of different images by using thresholding method. Finally, Support Vector Machine （SVM） was used to distinguish changed pixels from unchanged pixels in the multidi- mensional feature space. Experiment results show that the proposed method achieves better perform- ance than the KI threshold selection criterion based on Gaussian model （GM_KI）, KI threshold selec- tion criterion based on general Gaussian model（GGM_KI） and Otsu methods, the lowest Kappa is0.87, and the lowest anti-noise is 0.97 when the Peak Signal to Noise Ratio（PSNR） belongs to [29, 44]dB. These results verify the effectiveness and superiority of the proposed method.