《计算机应用研究》|Application Research of Computers

视频中遮挡情况下目标的跟踪

Video objects tracking under occlusions

免费全文下载 (已被下载 次)  
获取PDF全文
作者 梁华,刘云辉
机构 国防科学技术大学 电子科学与工程学院,长沙 410073
统计 摘要被查看 次,已被下载
文章编号 1001-3695(2008)03-0950-03
DOI 10.3969/j.issn.1001-3695.2008.03.088
摘要 提出了一种基于概率外观模型和Condensation的跟踪方法。该方法通过分析目标前景MBB重叠关系检测遮挡的产生和解除,在目标未遮挡时建立并更新目标的概率外观模型。当遮挡发生后,利用目标共面条件确定目标前后关系,通过合并目标模型计算系统观测似然度概率。算法统一在Condensation框架下进行有遮挡和无遮挡的跟踪。实验证明了该方法的有效性。
关键词 视频跟踪;遮挡;概率外观模型;Condensation
基金项目 国家自然科学基金资助项目(60334010,60475029,60675056);航天基金、武器装备预研基金资助项目(51416070205KG0173)
本文URL http://www.arocmag.com/article/1001-3695(2008)03-0950-03.html
英文标题 Video objects tracking under occlusions
作者英文名 LIANG Hua, LIU Yun-hui
机构英文名 College of Electronic Science & Engineering, National University of Defense Technology, Changsha 410073, China
英文摘要 This paper proposed a video objects tracking method based on probabilistic appearance models and Condensation. MBB overlapping features in successive frames were used to detect occlusions. Objects coplanar information was used to determine the objects order in merging. Probabilistic appearance models were overlaped to calculate the observation likelihood under occlusions. The tracking process was performed in a Condensation framework. Experiments of NUDT and PETS2001 video show the effectiveness of the proposed method.
英文关键词 video tracking; occlusions; probabilistic appearance model; Condensation
参考文献 查看稿件参考文献
 
收稿日期
修回日期
页码 950-952
中图分类号
文献标志码 A