一种基于分片模板的自适应运动目标跟踪算法

分块跟踪算法是一种非常鲁棒的模板匹配算法,特别适合解决部分遮挡问题,但是该算法框架下的模板更新策略无法解决场景中的目标外观变化和遮挡问题,为此,提出一种分块跟踪框架下的带有遮挡检测的模板更新算法。算法将目标分为多个小片,根据各小片匹配的情况检测目标是否被部分遮挡。如果目标被遮挡,使用能精确反映目标信息的参考模板进行匹配;否则使用能反映目标变化的临时模板,并提出了相应的模板更新算法。大量的实验证明了本算法的有效性。

Adaptive Object Tracking Algorithm Based on Fragment Template

Fragment tracking is a robust pattern matching algorithm.At present,the template update strategy of the fragment tracking algorithm is incapable under changed scenes,e.g.,target appearance changes,occlusions.Therefore,a template update method with the detection of occlusions was proposed.The target was represented by multiple image patches,and the proposed method distinguished occlusions according to the likelihood of the patches in target and the corresponding patches in the template.The well matched reference template was used for the next frame in the case of occlusions otherwise the temporary template which reflected the change of the target was used,and a corresponding template update method was proposed.Extensive experimental results on challenging sequence demonstrate the algorithm.