提出了一种通过直接分布函数密度比估计的SAR图像变化检测方法。不同于以往基于分布函数的方法需要先分别估计不同时相的分布函数再计算比值,该方法直接估计分布函数比值,并使用皮尔逊散度作为差异度量获取差异图。通过将概率标签松弛(PLR)嵌入到最大期望(EM)聚类算法中,增强了变化检测结果的空间一致性。实验结果表明,该方法能够准确地提取变化区域,得到较为满意的变化检测结果。

This paper introduces a new SAR image change detection approach based on density-ratio estimation. Unlike traditional distribution function based approaches which estimate distribution probability densities at different times respectively to calculate the ratio, our method directly estimates the ratio of the distribution probability densities and uses Person divergence as the measure of differences to generate difference map. By embedding the probability label relaxation (PLR) into the expectation maximization(EM) algorithm, the spatial consistency of the change detection results is enhanced. The experimental results show that our method can extract the changed areas accurately, and obtain satisfactory change detection results.

国家自然科学基金资助项目