基于速度场简化的人群行为分析

Crowd behavior analysis based on velocity field simplification

针对视频监控中人群异常行为难于被检测的问题，提出了一种基于速度场简化的人群行为分析方法。速度场简化方法利用拓扑简化和聚类算法对人群运动的速度场进行分析，提取出入群运动的宏观行为信息，之后利用这些信息检测人群中发生的各种异常事件。与其它方法相比，速度场简化方法具有不需要训练的优点，并且具有较好的鲁棒性。在买验中，利用速度场简化方法对人群整体行为异常、人群速度异常、人群间行为异常这三类事件进行了检测。实验结果表明，该方法的检测结果与真实情况十分接近。

To solve the problem that crowd abnormal behavior detection is difficult to be achieved during video surveillance, this paper proposes a crowd behavior analysis method based on velocity field simplification. The method of velocity field simplification uses topological simplification and clustering analysis to extract the information of a crowd, and then uses the information to detect the anomaly in the crowd. The biggest advantage of velocity field simplification method over others is that it does not need the time-consuming training. The proposed method was tested on the detection of three categories of abnormality： crowd formation/dispersal, changes in crowd speed and crowd splitting/merging. The experiments indicate that the proposed method can efficiently extract the main behaviors of crowd motion and the detection results are similar to the ground truth.