基于空间直方图的CamShift目标跟踪算法  

Object Tracking Using CamShift Algorithm Based on Spatiogram

作　　者：冀治航[1] 陈家新[1] 黎蔚[1] 杨凯鹏[1] 

机构地区：[1]河南科技大学电子信息工程学院,河南洛阳471003

出　　处：《微电子学与计算机》2009年第7期194-197,共4页Microelectronics & Computer

基　　金：国家自然科学基金项目(50571035)

摘　　要：为了提高在目标被遮挡及目标颜色发生较大变化情况下CamShift算法跟踪的稳健性,文中在目标建模阶段不仅利用了目标的颜色信息,而且还加入了目标的空间信息,建立了"空间直方图",从而提高算法的精确性及抗干扰能力;同时,依据径向距离理论提出了一种新的模板更新策略,进一步增强了算法的鲁棒性.实验结果证明,改进方法是有效的.Classic CamShift tracking algorithm limits the performance when the object is occluded or changes its color in large range. In order to improve the algorithm＇s stability, we introduce a new concept of a spatiogram, which includes not only the color information but also spatial information about the object. And using of the theory of radial-distance, we raise a new method to update the target model continually. Experimental results, tested on various sequence with scenes partial occlusion and rotations in depth, indicate that the two improvements enhance the precision and stability of the original algorithm.

关 键 词：CAMSHIFT 目标跟踪 空间直方图 径向距离 鲁棒性 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]