基于HOG特征的粒子滤波视频跟踪算法研究

Research on HOG-based particle filter vedio tracking

在视频分析处理领域中,特别是在视频监控领域,目标跟踪正在受到越来越多的关注。由于在实际应用中,利用运动摄像机拍摄的视频中,会造成背景的运动和目标尺寸的变化,即使是在固定摄像机拍摄的视频中,也会由于背景环境的复杂,造成目标的丢失和干扰。针对这一问题,为了改善在复杂场景下的目标跟踪效果,提出了结合梯度方向直方图(HOG)和粒子滤波的目标跟踪算法。此方法是通过在传统粒子滤波算法的算法框架下,增加目标跟踪的特征,提高了跟踪的鲁棒性,并根据检测结果确定目标。实验仿真表明,与传统单一特征的粒子滤波算法相比,文中的算法更能准确有效地跟踪复杂背景下的动态目标。

In the field of video analysis and processing, especially in video surveillance, target tracking is attracting more and more attention. In practical applications, video shot by motion cameras will cause background motion and target size change. Even if it is shot by a fixed camera, the video will also have target loss and interference due to the complexity of the background environment. In order to improve the effect of target tracking in complex scenes, this paper proposes a target tracking algorithm which combines the Histograms of Oriented Gradients （HOG） with the particle filter. This method increases the tracking features and improves the tracking robustness through the traditional particle filter algorithm. The target will be fixed based on testing results. The simulation result shows that this algorithm tracks the dynamic target more accurate and effective compared with the traditional particle filter algorithm with single feature.