目标窗口尺寸自适应变化的Mean-Shift跟踪算法

传统的窗宽尺寸固定不变的MeanShift跟踪算法不能实时地适应目标尺寸大小的变化。将多尺度空间理论与Kalman滤波器相结合，利用Kalman滤波器对尺寸变化的目标面积比例进行预测，用多尺度空间理论中的目标信息量度量方法求出前后相邻两帧的目标特征信息比，将其作为Kalman滤波器的观察值对目标面积比例进行修正，然后与MeanShift算法结合起来对目标进行跟踪，实验结果表明，改进的跟踪算法对尺度逐渐变大和变小的目标都能连续地自动地选择合适大小的跟踪窗口。

Mean-Shift tracking algorithm with adaptive bandwidth of target

The traditional Mean-Shift tracking algorithm of the fixed window-size cannot be adapted to real-time goal of the changes in size. Multi-scale space theory was combined with Kalman filter. First, Kalman filter was introduced to predict the proportion of the target image area, and then this proportion was revised by the observation, which was the proportion of the information of the two adjacent target images using the measurement of the target amount of information in the multscale space theory. Finally, it was implemented by the combination of the Mean-Shift tracking algorithm and Kalman filter to track targets. The improved algorithm can select the proper size of the tracking window in the scenarios that not only of increasing scale but of decreasing scale by the experimental results.