Research On Crowd Emotion Recognition And Contagion Method For Crowd Evacuation

Since entering the information age,the world economy has experienced rapid development.The density and mobility of crowds in public places have increased,leading to frequent public safety incidents.When emergencies occur in crowded places such as schools and shopping malls,panic and other negative emotions can spread among the crowd.This emotional contagion can result in overcrowding and stampedes,ultimately leading to safety accidents.By identifying the real emotions of the crowd,accurately calculating the level of panic within the group,and combining it with an emotion propagation model,it is possible to predict the number of people affected by panic accurately.This information can be used to regulate emotions,reduce panic among the masses,and ensure the safety of crowd evacuation.However,in real crowd evacuation scenarios,factors such as crowd obstruction can result in incomplete facial expression information,making it difficult to accurately identify group emotions.Additionally,traditional emotion contagion models rely on assumed parameters without real emotional data to support them.These models also do not consider the influence of group panic,deviating significantly from real-world crowd evacuation scenes.Therefore,accurately identifying individual emotions with incomplete facial expression information and integrating the calculated level of group panic into traditional emotion contagion models to predict changes in group emotions during evacuation remains highly challenging.To solve this problem,this thesis proposes a spatial-temporal consistent approach for accurate recognition of group emotions,considering the calculation of group panic in situations where facial expression information is incomplete.This method improves the accuracy of group emotion recognition and enables the development of an end-to-end crowd panic emotion recognition system.Furthermore,based on the proposed method,a group panic-based emotion propagation and evacuation simulation method is introduced.This method presents a GP-SIS emotion propagation model to explore the process of emotional contagion,making it more applicable to real-world crowd evacuation scenarios than traditional emotion propagation models.The main contributions and innovations of this thesis are as follows:(1)To solve the problem that existing group emotion identification methods is difficult to accurately identify crowd emotions in real crowd evacuation scenes due to crowd occlusion and other factors,this thesis propose a crowd emotion recognition method based on spatial-temporal consistency.The first step involves using the Resnet-GP network to accurately recognize individual emotion values in video frames that capture complete facial expression information.For frames with incomplete facial expression information,a spatial-temporal consistent model is proposed to calculate individual emotion values in each frame.The second step defines the level of panic within the crowd and calculates the real emotions of the group.The third step implements an end-to-end crowd panic emotion recognition system to validate the proposed method.Experimental results demonstrate that the spatial-temporal consistent approach accurately calculates the level of group panic,which is of significant importance in guiding crowd evacuation.(2)To solve the problem that the existing emotional contagion model does not consider the influencing factor of the degree of panic in the group,and it is difficult to simulate the emotional contagion process in the crowd evacuation scene,this thesis propose a crowd emotion contagion and evacuation simulation method based on group panic degree.The GP-SIS emotion contagion model incorporates the factor of group panic into the traditional SIS model to explore and analyze the process of emotional contagion,providing the number of infected individuals and the overall emotional contagion results of the entire group at different time points.Additionally,a small-world network with weighted group panic is constructed as a supplementary case to validate the effectiveness of the GP-SIS emotion propagation model.Experimental results show that the group panic-based emotion propagation and evacuation simulation method better aligns with the process of emotional contagion in real-world crowd evacuation scenarios.