视频跟踪算法研究综述  

Survey of Visual Tracking Algorithm

作　　者：闫庆森[1] 李临生[1] 徐晓峰[1] 王灿[1] 

机构地区：[1]太原科技大学电子信息工程学院,太原030024

出　　处：《计算机科学》2013年第06A期204-209,共6页Computer Science

基　　金：山西省研究生优秀创新基金项目(20123107)资助

摘　　要：在许多计算机视觉应用领域中,视频跟踪是最基本的任务。尽管有了大量的跟踪算法,但是跟踪算法的鲁棒性仍是具有挑战性的问题。物体的突然运动、目标或者背景外观的改变、目标与目标以及目标与背景的遮挡、非刚性物体的结构、摄像机抖动等问题都是视频跟踪算法设计过程中需要考虑的因素。介绍了视频跟踪算法及其研究进展,综述了现有基本的目标跟踪算法分类,详细描述了每种表示方法,并指出其优缺点。进一步讨论了跟踪的重要性问题,包括目标检测、特征选择、贝叶斯跟踪、在线学习跟踪等。Visual tracking is a fundamental task in many computer vision applications, and the robustness problem is still a challenge in spite of the numerous existing visual tracking algorithms. Besides, several circumstance, as the abrupt object motion, the variation of the target or the background, the object-to-object and obj ect-to-scene occlusions, the nonrigid object structures and camera jiggle, can reduce the effectiveness of the designed visual tracking algorithm. This survey described the visual tracking algorithm and its research advance. The existing target tracking algorithms were overviewed, and described each of these algorithms in detail, then, the advantage and disadvantage of each algorithm were analyzed respectively. Furthermore ： the important issues related to tracking were discussed, including detection of obiects, feature selection, Bayesian tracking, and online learning tracking.

关 键 词：视频跟踪 特征选择 贝叶斯跟踪 在线学习 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]