Mean Shift目标跟踪核函数宽度的自适应调整

Adaptive adjustment of kernel bandwidth for Mean Shift object tracking

Mean Shift算法无需穷尽搜索就可快速定位目标,因此被广泛应用于实时性要求较高的目标跟踪领域中。但传统Mean Shift算法的核函数宽度,也即跟踪窗口是固定的,不能适应目标大小变化,定位精度低。针对该问题,提出一种目标尺度度量方法,并应用于Mean Shift算法中,实现核函数宽度随着目标大小变化而自适应调整,实验仿真结果表明改进后的算法能很好地跟踪目标大小的变化,跟踪效果很好。

Mean Shift algorithm has been widely used in high real-time field of target tracking because it can converge quickly without exhaustive searching.However,the kernel bandwidth of this traditional algorithm is fixed,which can not adapt to the size change of the target or locate accurately.This paper proposes a target-scale method,which measures the scale of the target and applies it to Mean Shift algorithm in order to implement the kernel bandwidth self-adaptation according to the size of the target.The experimental results indicate that the kernel bandwidth can adapt to the size change of the target and the tracking result is improved.