基于谱残差法的主动声呐疑似目标自动拾取

Automatic pickup of suspected targets in active sonar basedon spectral residual method

针对主动声呐干扰遍布的B显画面上目标亮点难以拾取的问题,本文提出一种基于谱残差法视觉显著性处理的疑似目标自动拾取方法。在主动声呐图像的概念之下,研讨了自然图像中视觉显著体的发现与主动声呐图像中疑似目标拾取的相似机理,依此进而说明了本文方法的机理、核心框架以及关键处理环节,包括预处理声呐图像、生成显著图、估算拾取阈值以及标定疑似目标等。海试数据处理结果表明:相较于基于局部相对能量分析的对照方法,本文方法对已知目标具有更低的漏报率,同时亮点拾取数量减少约40%,且处理耗时减少约37%。

A method based on visual saliency spectral residual method is proposed for automatically picking up suspected targets out of complex and serious interferences background from the active sonar detecting output.After the conception of active sonar image built,the similarity between discovering visual saliency in natural images and picking up suspected targets in active sonar images is discussed.Based on this,the mechanism,core framework and key processing steps of the new method are explained,including preprocessing sonar images,generating saliency map,estimating pick-up threshold and picking up suspected targets.The results of sea trial data processing show that compared to the control method based on local relative energy analysis,the new method has a lower miss rate for known targets,and can reduce the pick-up number of bright spots by about 40%,which means a lower false alarm rate.Moreover,the processing time of this method is reduced by 37%approximately.