Font Size: a A A

Research On Crowd Grouping Algorithm Based On Individual Relationship

Posted on:2018-07-27Degree:MasterType:Thesis
Country:ChinaCandidate:G P LiuFull Text:PDF
GTID:2356330518968438Subject:Computer software and theory
Abstract/Summary:PDF Full Text Request
Since 21 st century,the rate of urbanization in countries around the world has been increasing steadily.A large population swarmed into the city,resulting in a significant increase in the density of the city's population density.In crowded public places,casualties accident caused by crowded and stampede are more and more frequent.Such public security problems have caused people panic among people,resulting in a large number of property losses,and also seriously interfere with the pace of urban development.Computer simulation technology provides a quick and secure way for crowd behavior simulation,which overcomes the drawbacks of using real people to simulate evacuation.Through the simulation of crowd behavior,when a public safety problem arises in a crowded public place,it can effectively carry out a scientific guidance for the huge crowd.In real life,people have the characteristics of conformity.The person who has the ability to act independently will be affected by the other people and the environment,in addition to the individual physiological and psychological factors.Especially in the process of crowd evacuation in face emergency situations,due to the reasons of the psychological driving force,close people will be gathered groups in a smaller range of space,resulting in the individual group behavior.It can effectively improve the crowd evacuation speed and improve the visual simulation effect by setting the appropriate inducing factors by using this behavior.In the present study,pedestrians are an isolated individual,and there is no association with the rest of the people around,and cannot really reflect the group characteristics in the movement process.Therefore,according to the above-mentioned problem,this paper proposes a K-Medoids clustering algorithm based on multi-impact factor to group the population,and then introduces relationship-based model of pedestrian social groups and applies it to the crowd evacuation simulation under virtual scene,and show good simulation results.The main work and innovation of this paper are as follows:1.Data extraction and analysis of crowd movement in real scene.In the square,teaching buildings and other crowded public places,different types of video capture devices are used to capture the crowd in the scene.And the captured video is used to analyze the crowd behavior.2.Propose a population grouping model based on relation and distance.In the crowd evacuation process,pedestrians will produce self-organization phenomenon,families,friends and other closely related people will be formed group according to the degree of intimacy.The closer the relationship between people,the higher their aggregation in the group,and the group behavior of the population existed throughout the evacuation process.Individuals in the crowd are affected by psychological factors besides physical location.In order to synthesize the real crowd behavior simulation results reflect the individual group behavior,the crowd movement simulation need to consider the relationship between the individual and distance on the impact of the group.In order to more realistic reflect the group in the movement of the phenomenon of grouping.3.Propose an improved clustering algorithm and apply it to crowd grouping.A K-Medoids clustering algorithm with multi-impact factor is proposed.The algorithm considers two different kinds of eigenvalues.In order to verify the effectiveness of the proposed algorithm,several experiments are designed.The experimental results show that this method can effectively improve the evacuation efficiency,and because the grouping process takes into account the distance between people and the degree of intimacy and other factors,can make crowd evacuation effect in the virtual environment closer to the crowd evacuation behavior in the real environment.4.Propose a relationship-based model of pedestrian social groups.Although the original social force model can simulate the export into arch,"fast is slow" and some other phenomenon.But in the movement process,pedestrian in the model is an isolated individual that is not associated with the surrounding neighbors.Therefore it can't truly reflect the group characteristics in the crowd movement.In order to simulate this behavior of pedestrians in the process of movement,a social groups force model was proposed.However,the social groups force model didn't take the influence of strength of group membership on the group behavior into account.In view of the above shortcomings,this paper proposes a pedestrian social groups force model based on kinship relationship strength to consider the influence of the relationship strength on the group attribution force in the model.The model can reflect the features of pedestrian behaviors in the process of movement,and achieve a more efficient and more realistic crowd movement.In order to verify the above theoretical research,the project team constructed the crowd modeling and motion simulation and photorealistic rendering platform according to the existing research theory.The crowd modeling and motion simulation platform can carry out the operation of scene import,semantic acquisition,crowd movement and so on.The rendering platform can simulate the movement of people under virtual scenes.Based on the two platforms,the influence of the group size,group relationship and total exit width of the scene on the evacuation time of the crowd is summarized by the simulation and analysis of the crowd motion of different complexity scenarios.The simulation results show that the algorithm model proposed in this paper can simulate the evacuation process and reflect the characteristics of the group,which has certain application value.
Keywords/Search Tags:Group behavior, Clustering algorithm, Crowd Evacuation Simulation, Social groups Force Model
PDF Full Text Request
Related items