自动选择跟踪窗尺度的Mean-Shift算法

实用的跟踪系统要求能实时地适应运动目标的外观变化,尺度固定不变的跟踪窗口不能有效地跟踪存在明显尺度变化的目标。本文将多尺度图像的信息量度量方法引入到运动目标跟踪中,提出了一种跟踪窗口自动更新算法,并用此算法改进了基于颜色直方图的Mean-Shift跟踪方案。实验结果表明,改进的跟踪算法对尺寸逐渐减小和逐渐增大的目标都能自动选择合适的跟踪窗口大小。

Mean Shift Tracking with Self-updating Tracking Window

A practical tracking system is required to update the appearance changes of moving objects in real-time.The system with fixed-size tracking window could not trace an object effectively when scale of the object has distinct changes,therefore it is important to select the scale of tracking window automatically.The information measure of multi-scale image in scale space has been used to differentiate the scale and is introduced into moving object tracking in this paper.Automatic updating method of tracking window is proposed,and is integrated into the classical Mean-Shift tracking algorithm based on color histogram.Experimental results demonstrat that the improved algorithm could select the proper size of the tracking window in the scenarios that not only of increasing scale but of decreasing scale.