摘要： 基于Mean Shift的目标跟踪算法，在目标发生明显尺度变化或存在背景干扰的情况下，跟踪就会失败。为此，针对跟踪过程中的背景干扰问题，提出根据目标运动状态进行背景滤波的目标跟踪算法。根据目标跟踪过程中产生的运动轨迹估计目标位移和速度，沿着目标可能的运动方向的反方向对候选区域进行背景滤波，滤波区域宽度根据目标位移大小确定。实验结果表明，改进后的算法对背景信息具有较好的鲁棒性，提高目标跟踪的可靠性。

关键词: 目标跟踪, 运动估计, 背景滤波, Mean Shift算法, 干扰

Abstract: Mean Shift based object tracking algorithm may fail while the target is changing in scale or in the interference environment. This paper presents a background filtering for object tracking according to the motion of target. The future displacement and speed of target according to its past moving track are estimated. Based on that, the target liked pixels is removed from the candidate location opposite the moving direction of target. The filter width is determined by the absolute value of target displacement. Experimental result shows the improved approach is robust against the interference background, and reliability of object tracking task is improved.

Key words: object tracking, motion estimation, background filtering, Mean Shift algorithm, interference

中图分类号: 

• TP391.41