基于改进CAMShift的运动目标跟踪算法

针对视频序列图像目标跟踪中Mean Shift算法提取目标颜色特征易受背景影响的问题,首先选取非线性核密度估计方法用来进行运动目标的检测,然后采用CAMShift方法对检测到的目标进行跟踪,并结合非线性核密度估计的检测结果对目标直方图进行自适应更新。还针对目标的遮挡问题给出解决方法。实验结果表明,引入背景减法与CAMShift相结合的策略,能够实现运动目标的自动跟踪,并实现目标直方图的自适应更新。该算法的可靠性能满足实时检测的要求,较好地解决了光照变化、阴影及遮挡等造成的影响。

Object tracking algorithm based on improved CAMShift

For the problem of Mean Shift algorithm extracting the target color features easily affected by the background, this paper selects non-linear kernel density estimation to detect single moving object, then uses CAMShift method to track the detected target, combined with test results of nonlinear kernel density estimation to do adaptive update of target histogram. This article also gives solution of target occlusion problem. Experimental results show that the method which combines the foreground detection and CAMShift can achieve automatic tracking of moving target, as well as adaptive update of target histogram. The reliable performance of this algorithm can satisfy real-time requirement, and eliminate the effects of unstable scene illumination and object occlusion.