摘要
视频监控系统在智能安防等众多领域有着广泛的应用空间,是近年来研究的热点之一。行人的分割在视频监控系统中极其重要,其分割的有效性直接影响目标识别、行为理解等。本文提出一种自动的行人分割算法,使用基于HOG特征的行人检测算法对目标进行定位,获得粗略的前景和背景信息,把获得的预分割信息作为GrabCut算法的先验知识,无需交互式操作,自动地对行人进行精确分割。实验证明该算法不仅有效,而且能满足应用需求。
Video surveillance system has been applied widely in many fields such as intelligent security, it is one of research fo- cuses in recent years. Segmentation of pedestrians in a video surveillance system is significant, and the effectiveness of the seg- mentation has a direct impact on the object recognition and behavior understanding. An automatic pedestrian segmentation algo- rithm is proposed in this paper. Firstly, the target is located based on the HOG pedestrian detection algorithm, and the foreground and background information is roughly obtained; then, the pre-segmentation information is used as the prior knowledge of GrabCut algorithm, without interactive operation, and the pedestrian is segmented automatically and precisely; Finally, the experimental results illustrate that the proposed algorithm is effective and meets the application requirements.
出处
《计算机与现代化》
2013年第8期7-10,共4页
Computer and Modernization
基金
国家自然科学基金资助项目(61073116
61202228)
