基于视觉显著性的雷达视频舰船检测  

Ship Detection in Radar Video Based on Visual Saliency

作　　者：周伟[1] 何东亮[1] 关键[1] 李国强[2] 

机构地区：[1]海军航空工程学院电子信息工程系,山东烟台264001 [2]火力控制技术国防科技重点实验室,河南洛阳471009

出　　处：《雷达科学与技术》2012年第1期54-58,63,共6页Radar Science and Technology

基　　金：国家自然科学基金(No.60672140;60802088);航空科学基金(No.20095184004);"泰山学者"建设工程专项经费资助

摘　　要：将计算机视觉领域的视觉显著性的概念引入常规雷达视频序列中舰船目标的检测,提出了适用于雷达视频的目标显著性表示模型。利用局部对比度来表征目标与杂波在回波强度上的差异,利用运动显著性来提取目标与固定地杂波和起伏海杂波之间的差异,经线性组合形成综合显著图,可快速准确提取目标。将连续多帧提取的结果进行积累,通过分析目标历史轨迹对目标加以确认。最后,利用采集的某型导航雷达实测数据进行的实验表明了方法的有效性。The concept of visual saliency in computer vision field was introduced to solve ship detection problem in general radar video sequence,in which a model was proposed to represent the saliency of target in radar.The local contrast was used to describe the difference of radar returns between target and clutter.The motion saliency was used to distinguish target from fixed land clutter and fluctuant sea clutter.Both these two features were linearly combined to form integral saliency map,in which targets can be extracted efficiently and correctly.The results from consecutive frames were integrated to construct target trajectory and target can be confirmed by its trajectory.Finally,tests on collected marine radar data verify the effect of the proposed algorithm.

关 键 词：雷达视频 舰船检测 视觉显著性 目标轨迹 

分 类 号：TN957.51[电子电信—信号与信息处理] TP391[电子电信—信息与通信工程]