Research On Early Warning Management Of Stampede Risk Based On Crowd Panic Monitoring

At present,urban public places are undertaking more and more activities in the fields of commerce,entertainment and culture,etc.Crowds are easily formed in these activities,causing the risk of stampede accidents in public places.Meanwhile,in the emergencies,such as fires,earthquakes,terrorist attacks,etc.,it's easy to cause mass gathering and crowd panic,which lead to secondary stampede risk.Thus,the early warning management of stampede risk is particularly necessary.The present studies on the early warning of stampede risk are mostly based on the monitoring of physical factors such as the crowd size,crowd density and abnormal behaviors,but few researchers focus on the monitoring of crowd panic.How to carry out the early warning management of stampede risk based on crowd panic monitoring and put forward the pre-control measures are new research topics.Considering both physical and psychological factors,this dissertation studies the early warning management of stampede risk from the perspective of crowd panic monitoring.The research content includes as following:(1)Utilizing the co-word analysis method,this dissertation conducts the encoding and clustering process of the keywords extracted from academic literatures and news.A panic chain model of stampede risk is built.The system and process for early warning management of stampede risk based on crowd panic monitoring are proposed.(2)Study the early warning process for the individual cause of stampede risk based on crowd panic monitoring.According to the face feature recognition and LDA linear discriminant method,this dissertation studies the identification process of individual cause characterization;the early warning model for the individual cause of stampede risk based on crowd panic monitoring is constructed and the dynamic switching of the identified targets is introduced.(3)Propose the early warning process for the crowd evolution of stampede risk based on crowd panic monitoring.The model of current crowd panic status is constructed.Based on the improved social force model and panic infection model,the crowd panic contagion is simulated and predicted.The early warning model of crowd panic evolution is proposed,which is adjusted based on false alarms and missed alarms.(4)The system of stampede risk pre-control is studied.It includes pre-control and site management.According to the early warning indicators of crowd panic monitoring,this dissertation proposes the strategies and implementation process for the precise pre-control.(5)Based on the early warning process and pre-control strategies,this dissertation utilizes the monitoring data,network data and simulation results to analyze the stampede accident in Shanghai Chen Yi Square and verify the early warning model.The innovation of the dissertation includes as following:(1)A three-dimensional system for early warning management of stampede risk based on crowd panic monitoring is proposed.The panic factors are used as main indicators,and the physical factors are used as auxiliary indicators to implement the early warning management process of stampede risk.(2)The dissertation proposes a facial recognition algorithm based on feature points and modeling,which realizes the identification of individual causes.(3)The early warning function of individual cause and crowd panic evolution is constructed,and the early warning process of stampede risk is proposed from the perspective of individual and crowd.(4)A risk-based pre-control system based on crowd panic early warning is studied.An accurate pre-control map for the early warning indicators is proposed.