基于置信值重构的视觉目标跟踪算法

Visual target tracking algorithm based on confidence value reconstruction

为了解决视觉目标跟踪过程中出现的目标表观变化和遮挡问题,在粒子滤波框架下,提出一种基于置信值重构的目标跟踪算法。通过对目标模板进行局部分块并结合提取的背景模板构建分类字典,利用概率协同表示分类算法,获得候选目标局部分块的分类概率。然后通过局部分块的分类概率重构候选目标的置信值。最终通过每个候选目标的置信值获得跟踪结果。实验表明,该文算法在目标表观变化和遮挡的情况下能够取得较好的跟踪效果。

In order to solve the problem of deformation and occlusion for the visual target tracking,a target tracking algorithm based on the confidence value reconstruction is proposed under the particle filter framework in this paper. A classification dictionary is constructed by combing the templates of background and the target＇s templates using the local patch method. Then,the local patch of the candidates is classified by the probabilistic collaborative representation,and their classification probabilities are further acquired. The confidence value of each candidate is constructed by using the classification probabilities of its local patches. Finally,the tracking result is obtained through the confidence value of all the candidates. Experimental results show that the proposed algorithm is effective for the visual tracking with the deformation and occlusion.