基于低秩约束的逆向联合稀疏跟踪算法

A Reverse Joint Sparse Object Tracking Algorithm via Low-rank Constraint

针对目标跟踪算法容易受到背景干扰的问题,提出一种基于低秩约束的逆向联合稀疏跟踪算法。该算法对目标模板和候选模板进行空间重叠分块,对目标模板施加低秩约束,降低杂乱背景的影响,再以候选模板构建视觉子字典联合稀疏表示候选模板。在粒子框架下采取三步打分法评价候选模板,并提出局部模板更新机制,以避免局部遮挡引起的跟踪漂移。实验结果表明,该算法在旋转、快速运动、遮挡、尺度变化和背景散乱等情况下均取得较好的跟踪性能。

The present object tracking methods are often interfered by scattered backgrounds,a reverse joint sparse object tracking algorithm via low-rank constraint was proposed to solve this problem.In the proposed method,the object templates were constructed by blocking processing,and then the low-rank constraints were imposed on these templates.After that,a three-step evaluation system was proposed in the particle framework.Then,a local template updating mechanism was adopted to avoid the occlusion interference.Experimental results verified that the method achieved good tracking performance in occlusion,rotation,motion blur,and scattered background.