摘要
目标跟踪是人工智能的研究热点之一。传统方法中,基于颜色直方图的目标跟踪易受背景相似颜色的影响。利用边缘方向直方图(EOH)方法对运动物体进行跟踪时,在复杂背景下其效果也会受到影响。文章对传统跟踪方法进行了改进,提出了一种利用拓扑模板进行跟踪的方法,对目标特征表示、参考模板更新部分做了相应改进。分块拓扑在保留传统跟踪方法对物体微小形变鲁棒性的同时,对被遮挡物体和形变物体有了更好的分辨能力。
Target tracking is one of the research focus of artificial intelligence.In the traditional method, the tracking target based on color histogram is easily affected by the similar background effects, edge orientation histogram （EOH） for moving objects tracking under complex background will also be affected.In this paper, the traditional tracing method is improved using a topology templates, improvements on feature representation and updating of target template are also made in this paper. Block topology in the traditional tracking methods not only keeps the robustness of small deformation, but also has better resolving power with the obscured objects and the deformation of targets.
出处
《信息网络安全》
2014年第1期52-55,共4页
Netinfo Security
基金
国家自然科学基金[60903065
61039003
61272052]
教育部博士点基金新教师基金[2009110212001]
关键词
目标跟踪
拓扑
模板
在线更新
object tracking
topology
template
online update
