Abstract: Automatic recognition of human behavior is an important but difficult problem in the area of computer vision. In this paper, a novel approach is introduced to handle the problem. Human body is detected using the contour information and the posture features are extracted by the contour fitting. A behavior classifier based on the normal behavior template is established to determine whether a human behavior is normal or not. Experimental results show that this system can run in real-time for the detection of abnormal behaviors with limited information and produce robust results by making full use of posture features information.

Key words: anomaly detection, posture feature, motion segmentation, video surveillance

摘要： 行人异常行为的自动检测与识别是计算机视觉领域的重点和难点，同时也是智能监控系统中研究的热点问题。针对这一问题，提出了一种基于人体形态特征的异常检测算法。利用轮廓信息将目标从视频序列中分割出来，再对分割出来的目标进行轮廓拟合，根据所得到的拟合信息提取文中所定义的形态特征因子，将特征因子经过行为分类器的判定，从而决策出该行为是否异常。实验结果表明该方法实现简单，具有较好的实时性与鲁棒性，可以作为实时监控系统中异常行为检测的有效方法。

关键词: 异常检测, 形态特征, 运动分割, 视频监控