基于CS的SAR图像自动目标分割算法

Automatic Target Segmentation in SAR Images Using CS

图像目标分割是SAR图像目标超分辨处理和自动目标识别的重要步骤。针对图像固有的稀疏结构,提出了一种SAR图像自动目标分割算法。通过构造变换字典将SAR图像数据投影到高维空间,实现了图像局部特征的稀疏表示,然后利用随机矩阵获得稀疏域局部特征的压缩采样,并对多组采样数据运用Mean-shift算法并行处理,最后通过符号检验法,实现了对目标像素与背景像素的分类。试验表明,该算法对硬目标具有较好的目标分割性能。

Object segmentation is an important step in SAR super-resolution processing and automatic target recognition. Considering image inherent sparse structures, an automatic target segmentation algorithm is proposed in this paper. First, a transformation matrix of dictionary is constructed to project the SAR image into a high dimensional space, and a sparse representation set of image local features is achieved. Second, a random sampling matrix is used to obtain its compression sampling and a mean-shift algorithm is applied to parallel process multiple sets of sample data. Finally, by using the sign test method, the SAR images data are classified as target pixels and background pixels classification. Experimental results demonstrate that the proposed algorithm has a good target segmentation results for hard target in synthetic aperture radar （SAR） images.