基于粒子滤波后验概率分布的多特征融合跟踪

在光照和目标形变等外部条件变化的情况下,仅利用目标的单一特征难以鲁棒的跟踪目标。提出了一种基于粒子滤波后验概率分布的多特征融合跟踪算法,在粒子滤波跟踪框架下,用直方图模型表征目标的颜色和边缘特征,通过两种特征后验概率之间的"协作"与"学习"实现特征融合,各种场景的试验结果比较表明,新的融合跟踪算法比仅用单一特征跟踪、现有的多特征融合算法具有更好的稳定性和鲁棒性,特别是针对环境光照和目标背景变化较大的情况更具有优势。

Fusing Multiple Features for Object Tracking Based on the posterior probabilities of Particle Filter

The object tracking only using single feature is apt to make errors or lose the target if the illumination and size scale-change. This paper presents a novel adaptive tracking algorithm that fuses multiple features based on posterior probabilities of the particle filter. In this paper uses histogram to characterize color and edge in particle filter algorithms framework. The posterior probabilities of all features are different from each other, which further promotes the fault tolerance ability by cooperation and learning from each other. An extensive number of comparative experiments show that the proposed tracking algorithm is more stable and robust than the single feature and other feature fusion tracking algorithms especially in the case of the illumination and background-change.