基于光流计算的异常拥挤行为检测算法

面向人群场景中异常拥挤行为检测,提出基于光流计算的检测方法。该方法首先采用光流微粒矢量场提取人群运动特征;然后基于社会力模型计算光流微粒之间的相互作用力;最后对相互作用力进行直方图熵值处理来实现人群行为判别。仿真实验表明,本算法可以区分人群场景中异常区域内相互作用力的大小,对异常拥挤行为进行判别和定位。

Abnormal crowd behavior detection based on optical flow computation

We propose a detection method based on optical flow to detect the abnormal crowded behavior in a crowd scene. In this method, the motion characteristics of the crowd are extracted by using the optical flow vector field. Then the interaction force between the particles is calculated based on the social force model. Finally, we implement histogram entropy discrimination process on the interaction force to achieve the discrimination of crowd behavior. Simulation results show that the algorithm can distinguish the size of the interaction force in abnormal regions in the crowd scene, identify and locate the abnormal congestion.