Mean shift算法作为一种非参密度估计算法，目前已被广泛应用于视频运动目标的跟踪。该算法具有运算效率快，对目标变形、旋转不敏感，在部分遮挡的情况下有一定鲁棒性等特点，但该算法在运动目标速度过快的情况下，由于没有考虑利用目标的运动方向和速度信息，因此在跟踪快速运动目标时容易造成跟踪丢失。针对此问题，提出了一种基于运动矢量分析与Mean shift跟踪算法相结合的新方法，即首先对视频编码过程中产生的运动矢量进行概率统计分析，以获取目标运动方向与运动速度的估计值，再以此修正Mean shift运动候选区域

Mean shift can make rapidly optimal matching during target tracking. But Mean shift algorithm doesnt use the targets motion direction and speed information. When targets moving speed is so fast that its easily fails to track the target accurately. We proposed a new target tracking algorithm combining motion vector analysis with Mean shift algorithm. By statistical analysis of the motion vector, we adjust the central point of the motion candidate region of Mean shift, which makes the starting position for search more close to the centre of the actual objective. New method not only improves the accuracy of fast moving tracking, but also reduces the iterative times during the center searching.