Research On Automatic Scenario Generation Method And The Crowd Behavior Analysis

In all kinds of large public places,there are a lot of security risks in crowd gathering activities.Public security and crisis management are facing unprecedented challenges.In order to effectively control and analyze the behavior of the population under dangerous conditions,it is necessary to analyze the relevant population behavior data.Because of the low frequency and great harm of dangerous events,the population analysis in dangerous situations is greatly restricted.Moreover,it is difficult for researchers to obtain data on crowd behavior in dangerous situations,which severely limits the research work on understanding and controlling of crowd behavior.This study takes analysis of crowd behavior in dangerous situation and uses crowd scenario generation technology to obtain crowd behavior data in virtual scenario by human-computer interaction,which effectively solves the problem of data acquisition.This research is mainly divided in the following two aspects.In the first aspect,we designed an interactive virtual crowd scenario generation method and built a high-simulation virtual crowd scenario which the obtained data can truly reflect the crowd behavior.In the second aspect,in the face of the data of population behavior in the virtual crowd scenario,we conducted in-depth research on data processing and analysis methods for revealing the potential rules of crowd behavior in typical scenarios.As a basic research in the field of public safety,this study provides theoretical and technical support for the public safety management of large-scale population.The main contributions of this paper are as following:(1)Starting from the design of virtual crowd scenario,this paper proposes a crowdsourcing data-driven method for the automatic scenario generation based on the generative adversial networks,which generates a series of real human-computer interaction scenario when the design requirements are met.(2)This paper takes the generated scenario data as the prototype and applies virtual reality modeling technology to propose a method of generating 3D virtual crowd scenario which is based on Unity3 D.In the meantime,it designs and implements a 3D virtual crowd simulation scenario.Focusing on the combination of environment,events and people,the research implements the rapid establishment of complex crowd simulation scenario,which can be used in various simulation experiments.(3)According to designing experiments and obtaining crowd behavior data,this paper proposes a fuzzy rule-based mining algorithm to model the decision-making behavior of people under dangerous conditions and realizes decision-making rules recognition driven by data.