Abstract: Change detection for SAR image can be achieved by clustering the difference image. Since SAR image can be interfered by speckle noise, thereby the change detection effect is affected. A method based on fuzzy clustering with spatial neighborhood information is proposed in this paper to improve clustering accuracy. According to the characteristics of the log-ratiomethod and the mean-ratio method, a new method is proposed to construct the difference image. This paper improves the traditional fuzzy clustering algorithm combined with the spatial neighborhood information of pixel to achieve change detection for SAR image. Compared with the traditional threshold algorithm, fuzzy C-means algorithm and fuzzy local information C-means algorithm, the experiments show that the proposed algorithm has high detection accuracy. It not only can effectively suppress the influence of noise, but also effectively retain the details of the image.

Key words: change detection, SAR image, fuzzy clustering, spatial neighborhood information

摘要： SAR图像变化检测可以通过对差异图的分类来实现，由于SAR图像容易受到相干斑噪声的干扰，从而影响变化检测效果。提出了一种基于空间邻域信息模糊聚类的SAR图像变化检测方法，根据对数比法和均值比法的各自特点，构造了一种新的差异图生成方法，并通过对传统的模糊聚类算法结合像素的空间邻域信息进行改进，来实现SAR图像的变化检测。实验结果表明，与传统的阈值法、模糊聚类算法以及局部邻域信息模糊C均值算法相比，提出的算法具有较高的检测精度，不但能有效地抑制噪声影响，同时能较好地保留图像细节信息。

关键词: 变化检测, SAR图像, 模糊聚类, 空间邻域信息