基于鲁棒主元分析的SAR图像目标分割

合成孔径雷达(SAR)图像的目标分割,是SAR图像自动目标识别的关键预处理步骤。与一般SAR图像目标区域分割方法不同,鲁棒主元分析融合了主元分析(PCA)与压缩感知(CS)理论中稀疏矩阵的先进思路,利用多帧具有相似性的SAR图像,构建一个观测矩阵D,通过求解一个凸优化问题,重建出一个低秩矩阵A和一个稀疏矩阵E。将矩阵A和E的列向量矩阵化,即可完成SAR图像目标与背景的分离。实验结果表明,鲁棒主元分析算法避免了复杂的SAR图像背景建模,针对同一目标的多帧SAR图像,所提方法对SAR图像目标和背景的分割问题具有可行性和有效性;与经典的最优阈值分割算法相比,误分率明显降低。

Target Segmentation of SAR Image via Robust Principal Component Analysis

Target segmentation from synthetic aperture radar(SAR) images is a critical preprocessing step for automatic target recognition in SAR images. Unlike general target region segmentation method of SAR images, robust principal component analysis(RPCA) combines the principal component analysis(PCA) and the advanced approach of sparse matrix in compressed sensing(CS) theory. With the similarity of multi-frame SAR images, an observation matrix D is built to solve a convex optimization problem, thus reconstructing a low-rank matrix A and a sparse matrix E. By matrixizing the each column of matrix A and E, target and background segmentation can be finished. The experimental results demonstrate that, when applied to multi-frame SAR images of the same target, RPCA avoiding the complicated background modeling of SAR image can achieve the segmentation of SAR image target and the background feasibly and effectively. Compared with the classical optimal threshold segmentation algorithm, the mis-segmentation rate is significantly reduced.