一种基于光流法的逆行异常事件检测方法

Method for Opposing Flow Abnormal Event Detection Based on Optical Flow

机场大厅、公路等公共场合,经常需要行人或车辆单向运动,以保障开放环境的安全性及秩序性。对监控视频中的逆行异常事件进行检测,便于管理人员及时对可疑事件做出处理。提出了一种逆行异常事件的检测算法,它基于图像中特征点的光流场以及空间分布特性对特征点进行聚类,然后通过计算符合逆行条件的特征点数量实现对逆行异常事件的检测。理论分析和实验结果均表明,该方法既能显著降低运算的复杂度,又能明显提高检测的准确率,适用于实时性要求较高的智能视频监控系统。

In many public places, such as airport lobby and high way, pedestrians and vehicles are required to run in one way traffic lane in order to guarantee the order and public security in open environments. It will be very convenient and effective for administers to handle urgent suspicious events if the surveillance system can automatically detect abnormal video events. A novel method for detecting opposing flow abnormal event is proposed, where optical flow and spatial distribution are utilized to cluster feature points in the frame, the number of feature points belonging to opposing flow is calculated to finally determine if there is an abnormal event. The theoretical analysis and experimental results show that the proposed method can dramatically reduce the complexity, and as well improve detection accuracy. It is quite essential for a real-time intelligent surveillance system.