依赖模板匹配纠错的混合跟踪算法*

经典均值漂移跟踪器(MST)和模板匹配跟踪器(TMT)在性能的许多方面具有强互补性。为利用这种互补性提高跟踪的可靠性，提出了一种混合跟踪算法(HTA)。HTA维持了颜色直方图和亮度模板两种目标模型，并分别以TMT和MST为主、辅跟踪器。当目标被遮挡或周围出现相似颜色模式的干扰物而造成主跟踪器错跟时，算法自动切换到辅跟踪器进行纠错；干扰消失或目标被重新捕获时，可自动切换到主跟踪器。实验结果表明，HTA对复杂跟踪环境适应性强且计算量不大。

Hybrid tracking algorithm recoverable from missing track by template matching

Mean shift tracker (MST) and template match tracker (TMT) have the total reversed characteristics in many aspects of tracking performance. In order to track object more reliable by using the complementarity of them, this paper proposed a hybrid tracking algorithm (HTA). HTA contains two target models, which was color histogram and intensity template, and used MST as primary tracker, TMT as auxiliary tracker. When primary tracker misses tracking probably caused by background clutter or total occlusion, HTA would switch to auxiliary tracker to remedy the mistake. When target is recaptured or clutter disappearances, primary tracker would be reactivated. Experimental results show the algorithm is robust to complex scene and has low computational cost.