在实时跟踪系统中,要求在跟踪过程中跟踪窗口的大小实时地适应运动目标的外观变化,这对应着Meanshift 的尺度变化. 针对跟踪这种尺度变化的问题,在跟踪框内检测角点并进行匹配,将所有得到的匹配角点对建立仿射模型,并采用最小二乘法求解得到尺度变化系数,进而更新跟踪框尺度,使得Meanshift 算法可自适应地变化尺度并跟踪到大小不断变化的目标. 实验结果表明,提出的算法具有较好的准确性、鲁棒性和实时性.-In real-time tracking system, requiring the tracking process to track the size of the window of moving objects in real time to adapt to changes in appearance, which corresponds to Meanshift scale changes. For the tracking problem of this scale changes in the tracking box corners and be detected match, will be matching all corners of the establishment of affine model, and using the least square method by scaling coefficient of variation, and then update the tracking box scale, making Meanshift algorithm can adaptively changing the scale and track the size of the ever-changing target. Experimental results show that the proposed algorithm has better accuracy, robustness and real-time.