Virtual crowd behavior modeling and parallel simulation method and system based on intelligent agent

Abstract

The invention belongs to the technical field of computer simulation, and provides a virtual crowd behavior modeling and parallel simulation method and system based on an intelligent agent. The method comprises the following steps: acquiring a plurality of monitoring video pairs of historical group activities, extracting first group feature data from the monitoring video pairs, constructing training data according to the first group feature data, and training a group behavior analysis model; receiving a third video of the real-time group activity shot by the unmanned aerial vehicle, extracting second group feature data and type information of the group activity from the third video, and matching according to the type information to obtain a plurality of abnormal events; performing deep analysis on the second group characteristic data and each abnormal event by using a group behavior analysis model to obtain a plurality of potential evacuation routes corresponding to each abnormal event, and performing parallel simulation on each potential evacuation route to screen out a target evacuation route; and outputting the target evacuation route to the management terminal. According to the invention, automatic generation of the evacuation route is realized based on the unmanned aerial vehicle remote photographing technology.

Description

Technical Field

[0001] The present invention relates to the field of computer simulation technology, and in particular to an agent-based virtual crowd behavior modeling and parallel simulation method and system. Background Art

[0002] With the development of society, group activities are increasing, such as large-scale concerts, sports events, commercial activities, etc. In these group activities, people are densely packed. Once an abnormal event occurs, such as fire, earthquake, etc., the evacuation of people becomes a crucial issue. In order to ensure that people can be evacuated safely in the shortest time, it is very necessary to plan a reasonable evacuation route in advance.

[0003] At present, in terms of evacuation route planning for group activities, managers manually set evacuation routes based on videos taken by drones, which is a current method. However, this method has many disadvantages. On the one hand, manually setting evacuation routes requires managers to spend a lot of time and energy to observe videos and analyze the distribution and activities of the crowd, which is extremely inefficient. On the other hand, due to the subjective judgment and experience limitations of managers, the manually set evacuation routes may not be completely reasonable and cannot fully take into account various complex situations, thus affecting the evacuation effect.

[0004] Therefore, there is an urgent need for a method that can efficiently and accurately plan evacuation routes for group activities to overcome the shortcomings of the existing technology. Summary of the invention

[0005] In order to solve the above technical problems, the present invention provides a method, system, electronic device, computer storage medium and computer program product for modeling and parallel simulation of virtual crowd behavior based on intelligent agents.

[0006] The present invention discloses a virtual crowd behavior modeling and parallel simulation method based on an intelligent agent, and the method comprises the following steps: obtaining a plurality of monitoring video pairs of historical group activities, wherein the monitoring video pairs comprise a first video and a second video, wherein the first video is a video before an abnormal event occurs, and the second video is a video after the occurrence of the same abnormal event until the evacuation is completed; extracting first group characteristic data from each of the monitoring video pairs, constructing training data based on the first group characteristic data, constructing a group behavior analysis model, and using the plurality of the training data for training; receiving a third video of real-time group activities shot by a drone, extracting second group characteristic data and type information of group activities from the third video, and obtaining a plurality of abnormal events based on matching of the type information; using the group behavior analysis model to perform in-depth analysis on the second group characteristic data and each of the abnormal events, obtaining a plurality of potential evacuation routes corresponding to each of the abnormal events, using simulation software to perform parallel simulation on each of the potential evacuation routes, and obtaining a target evacuation route based on screening of the simulation results; and outputting the target evacuation route to a management terminal of real-time group activities.

[0007] Optionally, extracting first group characteristic data from each of the surveillance video pairs and constructing training data based on the first group characteristic data includes: extracting first crowd behavior characteristics and building distribution information from the first video, extracting second crowd behavior characteristics and abnormal event types from the second video, and obtaining an evacuation effect evaluation value corresponding to the surveillance video pair; extracting evacuation channel distribution information and activity center area distribution information based on the building distribution information; taking the abnormal event type as the first label and the evacuation effect evaluation value as the second label, and constructing the first crowd behavior characteristics, the second crowd behavior characteristics, the evacuation channel distribution information, the activity center area distribution information, the first label, and the second label into a piece of training data.

[0008] Optionally, the extracting of the second group characteristic data from the third video includes: receiving a fourth video of real-time group activities taken by a drone, extracting appearance feature information and motion feature information of each human body from the fourth video, and obtaining a crowd diversity index based on a comprehensive evaluation of the appearance feature information and the motion feature information; determining a shooting duration based on the crowd diversity index, and controlling and receiving the third video of real-time group activities taken by a drone according to the shooting duration; and extracting the second group characteristic data from the third video.

[0009] Optionally, the method further includes: grouping each of the potential evacuation routes to obtain a plurality of evacuation route groups, and determining an execution time for each potential evacuation route in the evacuation route group; using simulation software to perform parallel simulation on each of the evacuation route groups, and in the process of simulating each of the evacuation route groups, overlappingly adding the potential evacuation routes in sequence according to the execution time, and evaluating the overall evacuation efficiency of each of the evacuation route groups; and determining the evacuation route group with the highest overall evacuation efficiency as the target evacuation route group.

[0010] Optionally, after outputting the target evacuation route to the management terminal of real-time group activities, the method further includes: the management terminal determines the control device and the guidance device corresponding to the target evacuation route, and after an abnormal event occurs, the management terminal controls each of the control devices and the guidance devices to be turned on together to perform an evacuation guidance operation.

[0011] The present invention also discloses a virtual crowd behavior modeling and parallel simulation system based on an intelligent agent, the system includes a processing device and a storage device, the computer program code stored in the storage device is executed by the processing device to achieve: obtaining a plurality of monitoring video pairs of historical group activities, the monitoring video pairs include a first video and a second video, the first video is a video before an abnormal event occurs, and the second video is a video after the same abnormal event occurs until the evacuation is completed; extracting first group feature data from each monitoring video pair, constructing training data based on the first group feature data, constructing a group behavior analysis model and using the plurality of the training data for training; receiving a third video of real-time group activities shot by a drone, extracting second group feature data and type information of group activities from the third video, matching and obtaining a plurality of abnormal events based on the type information; using the group behavior analysis model to perform in-depth analysis on the second group feature data and each abnormal event, obtaining a plurality of potential evacuation routes corresponding to each abnormal event, using simulation software to perform parallel simulation on each of the potential evacuation routes, and obtaining a target evacuation route based on the simulation results; and outputting the target evacuation route to a management terminal for real-time group activities.

[0012] The present invention also discloses an electronic device, comprising: at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, wherein the processor executes the computer program to implement any of the above methods.

[0013] The present invention also discloses a computer storage medium, wherein the computer-readable storage medium stores a computer program, and the computer program is executed by a processor to implement any of the above methods.

[0014] The present invention also discloses a computer program product, which includes computer codes. When the computer codes are executed by a processor of an electronic device, any of the above methods is implemented.

[0015] Compared with the traditional method of manually setting evacuation routes based on drone videos, the present invention greatly improves planning efficiency and avoids the time-consuming and labor-intensive human observation and analysis. At the same time, the analysis based on models and simulations effectively overcomes subjective judgment and experience limitations, can more comprehensively consider complex situations, make evacuation routes more scientific and reasonable, significantly improve evacuation efficiency and safety, and provide strong support for the safety of group activities. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0017] Figure 1 It is a flow chart of a method for modeling and parallel simulation of virtual crowd behavior based on intelligent agents disclosed in an embodiment of the present invention.

[0018] Figure 2 It is a structural diagram of an agent-based virtual crowd behavior modeling and parallel simulation system disclosed in an embodiment of the present invention. DETAILED DESCRIPTION

[0019] The following is a description of the implementation of the present application by specific specific embodiments. People familiar with the technology can easily understand other advantages and effects of the present application from the contents disclosed in this specification. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present application.

[0020] In addition, the technical features involved in the different embodiments of the present application described below can be combined with each other as long as they do not conflict with each other.

[0021] In response to the above technical issues, such as Figure 1As shown, an embodiment of the present invention discloses a method for modeling and parallel simulation of virtual crowd behavior based on intelligent agents, the method comprising the following steps: S101, obtaining surveillance video pairs of several historical group activities, the surveillance video pairs comprising a first video and a second video, the first video being a video before an abnormal event occurs, and the second video being a video after the same abnormal event occurs until evacuation is completed.

[0022] In this step, several pairs of surveillance videos of historical group activities are obtained, where each pair of surveillance videos contains two parts: the first video is a video taken just before an abnormal event (such as a fire or earthquake) occurs (for example, 3 minutes), which records the status of the group activity during normal progress, the distribution of personnel, the activity pattern, and other information; the second video is a video taken after the abnormal event occurs until the evacuation is completed, which records how the crowd evacuates after the abnormal event occurs and their behavior during the evacuation process.

[0023] S102, extracting first group characteristic data from each of the surveillance video pairs, constructing training data based on the first group characteristic data, constructing a group behavior analysis model and using some of the training data for training.

[0024] In this step, for each surveillance video pair, first group feature data can be extracted from the first video and the second video. The first group feature data includes crowd behavior features before and after the abnormal event occurs. Training data is constructed based on the extracted group feature data. The group behavior analysis model can be trained using the training data. The group behavior analysis model can be a model based on machine learning or deep learning, such as a neural network. The trained group behavior analysis model learns the behavior patterns and rules of the crowd in different states (normal and abnormal).

[0025] S103, receiving a third video of a real-time group activity shot by a drone, extracting second group feature data and type information of the group activity from the third video, and matching and obtaining a number of abnormal events according to the type information.

[0026] In this step, the real-time group activity is the target to be monitored by the present invention, and the deployed drone takes a video of the real-time group activity to obtain a third video. From the third video, the second group characteristic data can be extracted, and the second group characteristic data contains similar content to the first group characteristic data.

[0027] At the same time, the type information of group activities is extracted, such as concerts, sports events, exhibitions, etc. Different types of group activities may face different potential risks. For example, concerts may face risks such as fire and stampede, and sports events may face risks such as riots. According to the extracted type information, several possible abnormal events can be matched.

[0028] S104, using the group behavior analysis model to perform in-depth analysis on the second group characteristic data and each of the abnormal events, obtaining a number of potential evacuation routes corresponding to each of the abnormal events, using simulation software to perform parallel simulation on each of the potential evacuation routes, and screening out a target evacuation route based on the simulation results.

[0029] In this step, the trained group behavior analysis model is used to deeply analyze the second group feature data, and the evacuation behavior that the crowd may take corresponding to each abnormal event is predicted, so as to obtain several potential evacuation routes corresponding to each abnormal event. Then, these potential evacuation routes are simulated in parallel using simulation software (such as SUMO). During the simulation process, the crowd is simulated to evacuate along different potential evacuation routes, and various factors such as the passage capacity and crowd congestion are considered to obtain simulation results with corresponding evacuation efficiency evaluation information, and the potential evacuation route with the highest evacuation efficiency under various conditions is selected as the target evacuation route.

[0030] S105: Output the target evacuation route to a management terminal of real-time group activities.

[0031] In this step, the target evacuation route finally screened out is output to the management terminal of the real-time group activity, which can be the command center at the activity site, etc. The management personnel can formulate corresponding evacuation plans based on the target evacuation route. Once some abnormal event occurs, the crowd can be guided to evacuate more efficiently and accurately, thereby improving the evacuation effect and safety.

[0032] Compared with the traditional method of manually setting evacuation routes based on drone videos, the present invention greatly improves planning efficiency and avoids the time-consuming and labor-intensive human observation and analysis. At the same time, the analysis based on models and simulations effectively overcomes subjective judgment and experience limitations, can more comprehensively consider complex situations, make evacuation routes more scientific and reasonable, significantly improve evacuation efficiency and safety, and provide strong support for the safety of group activities.

[0033] An agent is a program or system with a certain degree of intelligence that can autonomously perceive, make decisions and take actions in a specific environment. The above scheme of the present invention can be executed by an agent, which can be embedded in a drone or deployed in a management terminal, and the present invention does not make specific limitations on this.

[0034] Optionally, extracting first group characteristic data from each of the surveillance video pairs, and constructing training data based on the first group characteristic data, includes: Extract the first crowd behavior characteristics and building distribution information from the first video, extract the second crowd behavior characteristics and abnormal event type from the second video, and obtain the evacuation effect evaluation value corresponding to the monitoring video pair; extract the evacuation channel distribution information and the distribution information of the activity center area according to the building distribution information; use the abnormal event type as the first label and the evacuation effect evaluation value as the second label, and construct the first crowd behavior characteristics, the second crowd behavior characteristics, the evacuation channel distribution information, the distribution information of the activity center area, the first label, and the second label into a training data.

[0035] In this embodiment, the first crowd behavior features are extracted from the first video before the abnormal event occurs, including the moving speed, moving direction, position, and aggregation degree of the crowd in a normal activity state, which reflects the behavior pattern of the crowd when it is not affected by the abnormal event. At the same time, the building distribution information is extracted, such as the layout of the building, the direction of the passage, etc.

[0036] The second crowd behavior features are extracted from the second video after the abnormal event occurs until the evacuation is completed, such as the change in the crowd's moving speed during the evacuation process, the choice of evacuation direction, whether congestion occurs, etc. These features show the crowd's behavioral response when facing abnormal events. In addition, the type of abnormal event is extracted, such as fire, earthquake, trampling, etc. The type of abnormal event is used to analyze the impact of different types of events on crowd behavior and evacuation effects. At the same time, the evacuation effect evaluation value corresponding to the monitoring video pair is obtained, which is obtained through personnel evaluation and relevant indicator calculation (such as evacuation time, number of evacuees, etc.), and is used to measure the effectiveness of this evacuation.

[0037] Based on the previously extracted building distribution information, we can further extract the evacuation channel distribution information, that is, determine the specific location, number, width, etc. of the evacuation channel, which is crucial for planning evacuation routes. At the same time, we can extract the distribution information of the activity center area, such as the location and size of the stadium's competition venue and the concert stage area at the event site. This information is conducive to better analyzing the gathering and flow of people in the event.

[0038] The above-mentioned first group behavior characteristics, second group behavior characteristics, evacuation channel distribution information, and activity center area distribution information (i.e., first group feature data) are used as the components of training data. At the same time, double labels are used, that is, the abnormal event type is used as the first label and the evacuation effect evaluation value is used as the second label. In this way, the group behavior analysis model obtained through training can learn the optimal planning method for evacuation routes for a certain type of abnormal event.

[0039] Optionally, the extracting of the second group characteristic data from the third video includes: receiving a fourth video of real-time group activities taken by a drone, extracting appearance feature information and motion feature information of each human body from the fourth video, and obtaining a crowd diversity index based on a comprehensive evaluation of the appearance feature information and the motion feature information; determining a shooting duration based on the crowd diversity index, and controlling and receiving the third video of real-time group activities taken by a drone according to the shooting duration; and extracting the second group characteristic data from the third video.

[0040] In this embodiment, the drone is first controlled to shoot a fourth video of real-time group activities, which is a sample video (the video length is, for example, 1 minute) for preliminary analysis of the basic characteristics of the crowd. The appearance feature information of each human body is identified from the fourth video, such as gender, approximate age range, body type (obese, medium, thin), clothing style, etc. These appearance features can be extracted by computer vision technology, such as face recognition, human posture estimation and other algorithms. At the same time, the motion feature information of each human body is extracted, including moving speed, moving direction, acceleration, etc. The motion feature information reflects the dynamic behavior characteristics of the crowd.

[0041] The crowd diversity index is an index used to comprehensively measure the degree of difference in crowd characteristics in real-time group activities. Based on the above appearance feature information and motion feature information, people with similar characteristics can be clustered. For example, people with consistent appearance features and consistent motion styles are classified into the same group, thereby obtaining the number of crowd types in the real-time group activity, and then obtaining the crowd diversity index. The crowd diversity index is, for example, an indicator value that is positively correlated with the number of crowd types, or it can be directly the number of crowd types.

[0042] When the number of crowd types is greater, it indicates that there are more types of participants in the real-time group activity, such as men, women, old and young, and the number of evacuation paths taken in the event of an abnormal event is greater. In other words, participants of the same type are more likely to choose similar evacuation paths (or escape paths). The more types of participants, the more escape paths are actually chosen. In order to accurately analyze and obtain potential evacuation routes that are closer to the actual situation, the present invention is configured to adjust the above-mentioned shooting duration according to the size of the above-mentioned crowd diversity index, that is, dynamically decide the duration of the third video to be shot. Specifically, the shooting duration is positively correlated with the crowd diversity index. When the third video is longer, more and richer second group feature data can be extracted from it, which is conducive to improving the accuracy of the potential evacuation route obtained by analysis. When the third video is shorter, the data length or data volume of the extracted second group feature data can be reduced, because the group feature data of highly similar participants (for example, almost all teenagers) is simple and obvious, and only a shorter third video is required to extract data that can fully characterize the group features, thereby reducing the data processing load.

[0043] In summary, the present invention first analyzes the fourth video to obtain the crowd diversity index, and then determines the shooting time to obtain the third video, which can more accurately extract the second group characteristic data and improve the accuracy and effectiveness of subsequent group behavior analysis and evacuation route planning.

[0044] It should be noted that the second group characteristic data in the present invention is different from the first group characteristic data, and it only includes the first group behavior characteristics, evacuation channel distribution information, and activity center area distribution information in the first group characteristic data.

[0045] Optionally, the method further includes: grouping each of the potential evacuation routes to obtain a plurality of evacuation route groups, and determining an execution time for each potential evacuation route in the evacuation route group; using simulation software to perform parallel simulation on each of the evacuation route groups, and in the process of simulating each of the evacuation route groups, overlappingly adding the potential evacuation routes in sequence according to the execution time, and evaluating the overall evacuation efficiency of each of the evacuation route groups; and determining the evacuation route group with the highest overall evacuation efficiency as the target evacuation route group.

[0046] In this embodiment, there are generally multiple evacuation channels (corresponding to an evacuation route) at the scene of the group activity. The traditional concept is that when an abnormal event occurs, opening all evacuation channels together will achieve the highest evacuation efficiency. However, at this time, participants will spend more time on the selection of multiple evacuation channels, resulting in the overall evacuation efficiency often decreasing instead of increasing. In response to this problem, the present invention is also configured to open multiple evacuation channels at the same time during evacuation, but the opening time of each evacuation channel is different. For example, when an abnormal event occurs, the first evacuation route is opened first, and then the second evacuation route is opened after a while, and so on. By gradually opening the evacuation routes, participants in group activities can be guided to evacuate more efficiently.

[0047] In order to achieve the above technical effects, the present invention divides each potential evacuation route into several evacuation route groups, the number of evacuation routes in each evacuation route group is different, and the execution time of each evacuation route is specified. Then, the simulation software simulates multiple evacuation route groups in parallel. For each evacuation route group, according to the above method, after reaching the corresponding time, a potential evacuation route is superimposed and opened. Multiple simulations can be performed on each evacuation route group. After the simulation is completed, all simulation results are integrated, and the overall evacuation efficiency of each evacuation route group is evaluated, and the evacuation route group with the highest overall evacuation efficiency is determined as the target evacuation route group.

[0048] Optionally, after outputting the target evacuation route to the management terminal of real-time group activities, the method further includes: the management terminal determines the control device and the guidance device corresponding to the target evacuation route, and after an abnormal event occurs, the management terminal controls each of the control devices and the guidance devices to be turned on together to perform an evacuation guidance operation.

[0049] In this embodiment, after receiving the target evacuation route (or target evacuation route group), the management terminal will determine the control devices and guidance devices located on the evacuation route according to the specific distribution of the route (or all routes in the target evacuation route group). The control device is, for example, a control system for a fire shutter door, which ensures that the evacuation passage is unobstructed by controlling the lifting and lowering of the shutter door; it can also be an elevator control system, which stops the elevator during evacuation to prevent people from being trapped; it can also be a ticket gate control system, which controls the opening of the ticket gate when an abnormal event occurs. Guidance equipment includes emergency indicator lights, voice broadcast systems, etc. The emergency indicator lights are usually installed on the ground or walls of the evacuation passage, and will light up after an abnormal event occurs to guide people to move towards the direction of the safe exit; the voice broadcast system can play evacuation instructions and precautions in real time, such as "Please don't panic, evacuate in an orderly manner according to the direction of the indicator light", etc., to help people stay calm and evacuate in an orderly manner.

[0050] After an abnormal event is detected, the management terminal will control all control devices and guidance devices to start together. The control device quickly adjusts the evacuation environment to create favorable conditions for personnel evacuation; the guidance device guides personnel to evacuate the scene in an orderly manner according to the target evacuation route through lighting and sound. The coordinated operation of various devices can maximize the evacuation efficiency and reduce casualties and property losses.

[0051] like Figure 2 As shown, an embodiment of the present invention further discloses a virtual crowd behavior modeling and parallel simulation system based on an intelligent agent, the system includes a processing device and a storage device, the computer program code stored in the storage device is executed by the processing device to achieve: obtaining a plurality of surveillance video pairs of historical group activities, the surveillance video pairs including a first video and a second video, the first video being a video before an abnormal event occurs, and the second video being a video after the same abnormal event occurs until the evacuation is completed; extracting first group feature data from each of the surveillance video pairs, constructing training data based on the first group feature data, constructing a group behavior analysis model and using the plurality of the training data for training; receiving a third video of real-time group activities shot by a drone, extracting second group feature data and type information of group activities from the third video, matching and obtaining a plurality of abnormal events based on the type information; using the group behavior analysis model to perform in-depth analysis on the second group feature data and each of the abnormal events, obtaining a plurality of potential evacuation routes corresponding to each of the abnormal events, using simulation software to perform parallel simulation on each of the potential evacuation routes, and obtaining a target evacuation route based on the simulation results; and outputting the target evacuation route to a management terminal for real-time group activities.

[0052] An embodiment of the present invention further discloses an electronic device, comprising: at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, wherein the processor executes the computer program to implement the method described in the above embodiment.

[0053] An embodiment of the present invention further discloses a computer storage medium, wherein the computer storage medium stores a computer program, and the computer program is executed by a processor to implement the method described in the above embodiment.

[0054] The computer-readable storage medium described above may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the above. Alternatively, the computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media may include electrical connections based on one or more lines, portable computer disks, hard disks, random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memory), optical fibers, portable compact disk read-only memories (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above.

[0055] An embodiment of the present invention further discloses a computer program product, which includes computer code. When the computer code is executed by a processor of an electronic device, the method described in the above embodiment is implemented.

[0056] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.

[0057] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for modeling and parallel simulation of virtual crowd behavior based on intelligent agents, characterized in that: The method comprises the following steps: obtaining a plurality of surveillance video pairs of historical group activities, wherein the surveillance video pairs comprise a first video and a second video, wherein the first video is a video before an abnormal event occurs, and the second video is a video after the occurrence of the same abnormal event until the evacuation is completed; extracting first group characteristic data from each of the surveillance video pairs, constructing training data based on the first group characteristic data, constructing a group behavior analysis model, and using the plurality of the training data for training; receiving a third video of real-time group activities shot by a drone, extracting second group characteristic data and type information of group activities from the third video, and obtaining a plurality of abnormal events based on matching of the type information; using the group behavior analysis model to perform in-depth analysis on the second group characteristic data and each of the abnormal events, obtaining a plurality of potential evacuation routes corresponding to each of the abnormal events, using simulation software to perform parallel simulation on each of the potential evacuation routes, and screening a target evacuation route based on the simulation results; The target evacuation route is output to a management terminal of real-time group activities.

2. The method for modeling and parallel simulation of virtual crowd behavior based on an intelligent agent according to claim 1, characterized in that: The first group characteristic data is extracted from each of the surveillance video pairs, and training data is constructed based on the first group characteristic data, including: extracting first crowd behavior characteristics and building distribution information from the first video, extracting second crowd behavior characteristics and abnormal event types from the second video, and obtaining an evacuation effect evaluation value corresponding to the surveillance video pair; extracting evacuation channel distribution information and activity center area distribution information based on the building distribution information; taking the abnormal event type as the first label and the evacuation effect evaluation value as the second label, and constructing the first crowd behavior characteristics, the second crowd behavior characteristics, the evacuation channel distribution information, the activity center area distribution information, the first label, and the second label into a piece of training data.

3. The method for virtual crowd behavior modeling and parallel simulation based on an agent according to claim 2, characterized in that: The extracting of the second group characteristic data from the third video includes: receiving a fourth video of real-time group activities shot by a drone, extracting appearance characteristic information and motion characteristic information of each human body from the fourth video, and obtaining a crowd diversity index based on a comprehensive evaluation of the appearance characteristic information and the motion characteristic information; determining a shooting duration based on the crowd diversity index, and controlling and receiving the third video of real-time group activities shot by a drone according to the shooting duration; and extracting the second group characteristic data from the third video.

4. The method for modeling and parallel simulation of virtual crowd behavior based on an agent according to claim 1, characterized in that: The method further includes: grouping each of the potential evacuation routes to obtain a plurality of evacuation route groups, and determining an execution time for each potential evacuation route in the evacuation route group; using simulation software to perform parallel simulation on each of the evacuation route groups, and in the process of simulating each of the evacuation route groups, sequentially and overlappingly adding the potential evacuation routes according to the execution time, and evaluating the overall evacuation efficiency of each of the evacuation route groups; and determining the evacuation route group with the highest overall evacuation efficiency as the target evacuation route group.

5. The method for virtual crowd behavior modeling and parallel simulation based on an agent according to claim 1, characterized in that: After the target evacuation route is output to the management terminal of real-time group activities, the method further includes: the management terminal determines the control device and the guidance device corresponding to the target evacuation route, and after an abnormal event occurs, the management terminal controls each of the control devices and the guidance devices to be turned on together to perform an evacuation guidance operation.

6. An agent-based virtual crowd behavior modeling and parallel simulation system, characterized by: The system includes a processing device and a storage device, and the computer program code stored in the storage device is executed by the processing device to achieve: obtaining a plurality of surveillance video pairs of historical group activities, wherein the surveillance video pairs include a first video and a second video, wherein the first video is a video before an abnormal event occurs, and the second video is a video after the same abnormal event occurs until evacuation is completed; extracting first group feature data from each of the surveillance video pairs, constructing training data based on the first group feature data, constructing a group behavior analysis model, and using the plurality of the training data for training; receiving a third video of real-time group activities shot by a drone, extracting second group feature data and type information of group activities from the third video, and matching a plurality of abnormal events based on the type information; using the group behavior analysis model to perform in-depth analysis on the second group feature data and each of the abnormal events, obtaining a plurality of potential evacuation routes corresponding to each of the abnormal events, using simulation software to perform parallel simulation on each of the potential evacuation routes, and screening a target evacuation route based on the simulation results; The target evacuation route is output to a management terminal of real-time group activities.

7. The agent-based virtual crowd behavior modeling and parallel simulation system according to claim 6, characterized in that: The first group characteristic data is extracted from each of the surveillance video pairs, and training data is constructed based on the first group characteristic data, including: extracting first crowd behavior characteristics and building distribution information from the first video, extracting second crowd behavior characteristics and abnormal event types from the second video, and obtaining an evacuation effect evaluation value corresponding to the surveillance video pair; extracting evacuation channel distribution information and activity center area distribution information based on the building distribution information; taking the abnormal event type as the first label and the evacuation effect evaluation value as the second label, and constructing the first crowd behavior characteristics, the second crowd behavior characteristics, the evacuation channel distribution information, the activity center area distribution information, the first label, and the second label into a piece of training data.

8. An electronic device comprising: At least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, wherein the processor executes the computer program to implement the method according to any one of claims 1 to 5.

9. A computer storage medium storing a computer program, characterized in that: The computer program is executed by a processor to implement the method according to any one of claims 1 to 5.

10. A computer program product, characterized in that: The computer program product includes computer codes, and when the computer codes are executed by a processor of an electronic device, the method according to any one of claims 1 to 5 is implemented.

Images