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.