基于粒子滤波的视觉目标跟踪算法

Visual Target Tracking Algorithm Based on Particle Filter

摘要: 随着社会智能化的发展，视觉目标跟踪成为计算机视觉领域的研究热点之一。在目标跟踪过程中，由于目标自身及环境的变化使得准确跟踪目标变得十分困难。在基本粒子滤波框架下，本文主要研究了一种基于颜色特征的粒子滤波视觉目标跟踪算法。通过引入核函数的RGB颜色空间对目标进行鲁棒的表达，为了适应跟踪过程中的目标变化，利用实时的观测信息对目标模板进行更新。实验表明，基于颜色特征的粒子滤波算法对光照变化和动态干扰具有较强的鲁棒性。

Abstract: With the development of social intelligence, visual target tracking has become one of the research hotspots in the field of computer vision. In the process of target tracking, it becomes very difficult to track the target accurately because of the change of the target itself and the environment. In the basic particle filter framework, this paper mainly studies a particle filter visual target tracking algorithm based on color features. By introducing the RGB color space of kernel function to the robust expression of the target, in order to adapt to the target change in the tracking process, the target template is updated with real-time observation information. Experiments show that the particle filter algorithm based on color features has strong robustness to light changes and dynamic interference.