Auto-detection of crowd behavior changes in lager assemblies

A primary concern in the field of automatic recognition of crowd behavior is to construct a universal crowd mutation recognition model. The Shannon entropy model was used to create a crowd status entropy model by targeting the assembly occupants macro order status and identifying individual speeds and related probabilities. The entropy model was validated through simulations and video detection. The results show that when the crowd macro behavior suddenly changes (for example from the disordered to fully-ordered status, or from the fully-ordered to the disordered status), the crowd macro status entropy changes. Thus, the entropy changes can be used to automatically detect the crowd behavior changes and to set warning alarms.