基于显著性检测的合成孔径雷达目标检测算法

Synthetic aperture radar target detection algorithm based on saliency detection

针对合成孔径雷达图像边缘、纹理与形状信息丰富,图像尺度大难以处理的问题,提出了一种基于显著性检测的合成孔径雷达目标检测算法.将合成孔径雷达图像表示成超像素,来构建超像素间的关联关系图从而解耦整张图像.同时基于人眼视觉系统对多尺度图像进行非均匀采样的特点,分别提取图像的局部对比度显著性、像素紧凑度显著性和全局唯一性显著性映射,并使用贝叶斯估计得到SAR图像的精确显著性映射,融合三种显著性映射得到最终的显著性图实现目标检测.在各种SAR图像显著性检测实验的定性与定量结果表明,所提方法明显优于现有方法.

Aiming at the problems concerning the edge,texture and shape information of the synthetic aperture radar image and the large image scale,a synthetic aperture radar target detection algorithm based on saliency detection was proposed.The synthetic aperture radar image was expressed as super pixels to construct the correlation diagram among super pixels to decouple the entire image.At the same time,based on the characteristics of non-uniform sampling of multi-scale images by the human visual system,the local contrast saliency,pixel compactness saliency and global unique saliency mappings were extracted,respectively.Bayesian estimation was used to obtain the pixel-level accurate saliency maps of the SAR image,and to merge the three saliency maps for the final saliency map and target detection.The qualitative and quantitative results of various SAR image saliency detection experiments show that the as-proposed method is significantly better than the currently existing methods.