基于背景优化的Mean Shift目标跟踪算法

Mean Shift tracking algorithm based on background optimization

针对传统的Mean Shift算法在目标快速运动且背景区域变化较大时,容易丢失跟踪目标的问题,提出了一种基于背景优化的Mean Shift目标跟踪算法。该算法引入混合直方图并对直方图重新量化,再通过减少背景像素在概率密度函数(PDF)中的权重来对背景进行优化,从而降低背景区域对跟踪的影响。实验结果表明,当目标快速运动,且背景区域变化较大时,该算法仍然能够实现对运动目标的准确跟踪。

The traditional Mean Shift tracking algorithm would lose target when the target is in rapid movement or the background is changing obviously. A Mean Shift tracking algorithm based on background optimization was proposed. The algorithm used cross-join histogram and re-quantilization, and optimized background by decreasing the weight of the Probability Density Function （PDF） of the background pixels. Then the impact of the background on the target region was reduced. The experimental results show that, when the target moves rapidly and the background region changes a lot, the proposed method can still track the target accurately.