时空上下文与CamShift相结合的目标跟踪算法

针对传统的时空上下文(STC)目标跟踪算法在完全遮挡或者遮挡面积过大时易导致跟踪失败的问题,提出了一种将STC与CamShift相结合的目标跟踪算法.通过设定一个阈值,来判断时空上下文算法何时进入目标遮挡.当进入遮挡时,利用CamShift算法得到的跟踪中心修正时空上下文模型计算出的跟踪中心,并用修正后的中心更新局部上下文区域.实验结果表明:提出的算法较原有的算法更加适合复杂的场景变化,具有更好的鲁棒性和稳定性.

Target tracking algorithm combines STC with CamShift

Aiming at the problem that the conventional spatio-temporal context(STC)target tracking algorithm can lead to tracking failure when total occlusion or overshooting area is too large,a target tracking algorithm which combines STC with CamShift is proposed.By setting a threshold,determine when the STC algorithm detects the target occlusion.When entering the occlusion,the tracking center obtained by the CamShift algorithm is used to correct the tracking center calculated by the STC model and update local context area with the revised center. Experimental results show that the proposed algorithm is more robust and stable than the original one in adapting to more complicated scene changes.