基于多表观模型竞争的视觉跟踪算法

Visual tracking algorithm based on competition of multiple appearance models

在视觉跟踪中,如何适时地更新目标模型是影响跟踪算法跟踪精度和鲁棒性的关键性因素,也是当前研究中面临的重点和难点问题。对此,提出了一种基于多表观模型竞争的模型更新策略。通过多表观模型中各子模型的贡献度大小确定竞争优势排序,当最优子模型的贡献度满足多表观模型更新阈值时,对各子模型及其对应的系数进行更新;否则,仅对部分子模型进行更新。在此基础上,以粒子滤波算法为跟踪框架,提出了基于多表观模型竞争的视觉跟踪算法。实验结果表明,所提算法能够较好地处理视觉跟踪中的模型更新问题,跟踪性能较无模型更新策略的粒子滤波算法有明显提高。

During the visual tracking, how to update the target model timely is the key factor that affects the tracking perfor- mance and robustness. This problem is also considered to be the important and hard issue of the visual tracking. To this point, this paper proposed a model update strategy based on the competition of multiple appearance models. It assigned the competition superiority order according to the contribution of each sub model. When the contribution of the best sub model subjects to model update threshold, it updated all the sub models and the corresponding weights. Otherwise, it partially updated the multiple appearance models. Based on this, it proposed a visual tracking algorithm based on multiple appearance models under the framework of particle filter. The experimental results indicate that the proposed algorithm can well cope with the model update problem, which obviously improves the tracking performance compared with the particle filter tracking without model update.