Crowd Status Analysis And Abnormal Behavior Detection

The accidents of endangering public safety occur frequently,such as a stampede?aggregation. Analysis of crowd status and abnormal behavior detection has important practical significance. The study of human behavior can provide important guidance information about the crowd management and scheduling.Harris corner acts as crowd feature points, the crowd motion vector field is obtained by using optical flow method to track the crowd feature points. The crowd motion intensity index is defined by crowd motion vector field and theorem of kinetic energy to describe the crowd movement intensity. The direction distribution of crowd is obtained by crowd motion vector field,furthermore, the crowd chaos index is defined by entropy to describe direction of crowd movement. The crowd status index is defined by the crowd movement area accounted for the proportion of the crowd foreground. The experimental results show that the three crowd status description operator can be a good description of the crowd movement status.As the crowd abnormal behavior detection algorithms are with high complexity and large amount of computation, it is difficult to ensure the real-time. A hierarchical processing crowd abnormal behavior detection scheme is proposed by the three descriptions of crowd operators. The experimental results show that the scheme has better effect and reduce the computation and improve the real-time performance.The gathered crowd is easily accompanied by violent uncontrollable events, but little attention has been the crowd gathers behavior. A model of crowd gather and diffusion is proposed based on characteristics of grid units, experimental results show that the model can be a good description of crowd agglomeration.