基于视觉注意力计算的运动目标检测方法研究

摘要

为了更准确地在全局运动视频场景中检测运动目标，提出了一种基于运动注意力和粒子滤波自底向上和自顶向下相结合的运动目标检测方法。基于多尺度可变块运动估计估计运动矢量场（Motion Vector Filed，MVF），构建运动注意力模型，得到运动注意力显著图，继而得到运动注意力的初始分布；采用自顶向下的基于目标颜色信息的粒子滤波算法，调整运动注意力的分布状况；使注意力集中到待测目标上，并提取出待测运动目标。实验结果表明，该方法在全局运动场景中能更加准确地检测目标。%In order to detect moving target more accurately in global motion scene, a moving target detection method based on the motion attention and particle filter combining bottom-up and top-down is proposed. The motion attention model is built based on Motion Vector Field(MVF)estimated by multi-scale variable block motion estimation, the motion attention map and the initial distribution of the motion attention are obtained. A top-down particle filter with color infor-mation is adopted to adjust the distribution of motion attention. Attention only focuses on the expected moving target, which is convenient for target extraction. The experimental results show that the proposed method can detect target more accurately in the global motion scene.