摘要
在计算机视觉领域,人体运动分析的研究正因其广泛的应用前景而越来越受到研究者的重视。对于单目图像序列的人体运动跟踪,现有的方法大多需要进行人工干预,或者在身体上附着标志物,或者对第一帧图像进行手工标记。本文针对这一问题,提出一种简单而有效的自动检测人体腿部骨架的算法,该算法不需要任何人工干预,继而在传统的运动建模及矩形块RGB颜色匹配的基础上,提出一种圆周相交定点算法,结合踝关节的运动预测对获取的腿部骨架进行跟踪,有效地解决了两腿的自遮挡问题。论文最后给出了跟踪的实验结果。
In recent years,the interest in human motion analysis has increased rapidly, due mostly to the large number of potential applications for this technology. With regard to the human motion tracking in monocular video sequences,manual intervention is often required in current methods,either attatching markers to the moving bodies,or marking the feature points manually in the initial frames.To solve this problem, this paper presents a simple but effective method for automatically extracting human leg skeletons,without any manual intervention. Furthermore,we track the leg skeleton by circle intersection pointing and the ankle motion prediction based on motion modeling and retangular region matching according to its colour,by which the problem of self-occlusion can be solved effectively.The final part of the paper lists the experemental results.
出处
《计算机工程与科学》
CSCD
2007年第1期62-65,共4页
Computer Engineering & Science
基金
国家自然科学基金资助项目(60573079)
湖南省自然科学基金资助项目(03jjy6025)
湖南省教育厅科研资助项目(03C227)
关键词
图像序列
人体运动
运动预测
匹配
腿部骨架
video sequences
human moton
motion prediction
matching
leg skeleton
