Behavior anomaly identification method, device and system during crowd gathering in specific occasion

Abstract

The embodiment of the invention provides a behavior anomaly identification method, device and system during crowd gathering in a specific occasion, and relates to the technical field of artificial intelligence, and the method comprises the steps: obtaining the crowd density of a crowd when detecting that crowd gathering exists in the specific occasion; judging whether the crowd density reaches a preset crowd density threshold value or not, if yes, judging that crowd density exists, and at the moment, obtaining the correlation degree of the crowd; acquiring a foreign person proportion in the crowd, and acquiring a time factor and a movement track of the crowd; and inputting the crowd density, the association degree, the foreign person proportion, the time factor and the action track into a pre-trained anomaly deep learning model, and identifying the anomaly of the crowd through the anomaly deep learning model. According to the embodiment of the invention, when a crowd gathering condition exists in a specific occasion, the current abnormal degree (such as crowd conflicts, sudden diseases, normal gathering and the like) of the crowd can be judged, and preparation is made for taking corresponding processing measures next.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a method, device and system for identifying abnormal behavior when a crowd gathers in a specific situation. Background Art

[0002] With the acceleration of urbanization and the increasing frequency of social activities, public safety issues are receiving more and more attention. Especially in specific places such as industrial parks, large gatherings, transportation hubs and commercial centers, various safety accidents are prone to occur when people gather. Although traditional monitoring systems can provide real-time video monitoring and then judge the behavior of the crowd based on the monitoring content, this method relies on manual observation and judgment, and has problems such as slow response speed, high false alarm rate, and difficulty in processing large-scale data.

[0003] The existing technology uses the following methods to judge the current behavior of the crowd in a specific occasion, so as to take corresponding treatment measures and improve the level of public safety. Method 1, a method based on machine learning rules: detect abnormal behavior of the crowd through predefined rules. For example, set a threshold for the density of people in a certain area, and trigger an alarm when the threshold is exceeded. Method 2, a method based on deep learning statistics: statistically model a large amount of normal behavior data, establish a normal behavior model, and compare it with the current behavior to identify abnormal behavior. However, both methods have defects. Although method 1 is simple and intuitive, it lacks flexibility and is difficult to cope with complex and changeable actual scenarios; although method 2 can adapt to a certain degree of change, it has high requirements for data volume and quality, and has limited generalization ability in new environments. In addition, although the above methods have improved the ability of abnormal behavior detection to a certain extent, they still have the following shortcomings: ① Poor environmental adaptability: The behavior patterns of people in different occasions and environments vary greatly. Existing methods are often difficult to adapt quickly in new environments, resulting in high false alarm rates and missed alarm rates. ② Difficulty in identifying complex behaviors: Some complex behaviors in crowd gatherings are difficult to effectively identify through simple rules or statistical methods. For example, how to judge the difference between group conflicts, sudden illnesses, and normal gatherings.

[0004] Therefore, the prior art has defects and needs to be improved and developed. Summary of the invention

[0005] The embodiments of the present application provide a method, device and system for identifying abnormal behavior of a crowd when they gather in a specific occasion. When a crowd gathers in a specific occasion, the current abnormality of the crowd (such as group conflict, sudden illness, normal gathering, etc.) can be determined.

[0006] The present application provides a method for identifying abnormal behavior when a crowd gathers in a specific occasion, the method comprising:

[0007] When a crowd is detected to be gathered in a specific occasion, obtaining the crowd density of the crowd;

[0008] Determine whether the crowd density reaches a preset crowd density threshold, and if so, determine that there is a crowd density condition, and obtain the correlation degree of the crowd;

[0009] Obtaining the proportion of outsiders in the population, as well as obtaining the time factor and the movement trajectory of the population;

[0010] The crowd density, correlation degree, proportion of outsiders, time factor and action trajectory are input into a pre-trained abnormality deep learning model, and the abnormality of the crowd is identified through the abnormality deep learning model.

[0011] In the method for identifying abnormal behavior when a crowd gathers in a specific occasion described in the embodiment of the present application, the degree of association includes a degree of familiarity, and obtaining the degree of association of the crowd includes:

[0012] Obtain the number of first persons in the group of people who belong to the same company, and obtain the number of second persons in the group of people whose company belongs to the same floor;

[0013] Get the total number of people in the group;

[0014] The familiarity level is generated based on the first number of people, the second number of people, and the total number of people.

[0015] In the method for identifying abnormal behavior when a crowd gathers in a specific occasion described in the embodiment of the present application, before obtaining the first number of people in the crowd who belong to the same company and obtaining the second number of people in the crowd whose company belongs to the same floor, the method further includes:

[0016] The basic information of each person in the crowd is obtained, wherein the basic information includes the company to which the person belongs and the floor on which the person's company is located.

[0017] In the method for identifying abnormal behavior when a crowd gathers in a specific occasion described in the embodiment of the present application, the degree of association includes interaction intensity, and obtaining the degree of association of the crowd includes:

[0018] Obtaining facial expressions, voice information, and body movements of each person in the group;

[0019] The facial expressions, voice information and body movements are input into a pre-trained deep learning model, and the interaction intensity of the group of people is generated by the deep learning model.

[0020] In the method for identifying abnormal behavior of a crowd gathering in a specific occasion described in the embodiment of the present application, the abnormality includes normal and abnormal. After identifying the abnormality of the crowd, the method further includes:

[0021] If the abnormality is abnormal, corresponding processing measures are triggered based on the abnormal situation.

[0022] In the method for identifying abnormal behavior when a crowd gathers in a specific occasion described in the embodiment of the present application, obtaining the proportion of outsiders in the crowd includes:

[0023] Get the total number of people;

[0024] Get the number of foreigners in the population;

[0025] The foreign personnel ratio is generated based on the number of the foreign personnel and the total number of the population.

[0026] In the method for identifying abnormal behavior when a crowd gathers in a specific occasion described in the embodiment of the present application, obtaining the crowd density of the crowd includes:

[0027] Obtain the area of ​​the specific occasion;

[0028] Get the total number of people in the specific situation;

[0029] The crowd density is generated based on the area of ​​the region and the total number of people.

[0030] An embodiment of the present application also provides a device for identifying the degree of abnormal behavior when a crowd gathers in a specific occasion. The device includes a processor and a memory. The memory stores a computer program. The processor calls the computer program stored in the memory to execute the method for identifying the degree of abnormal behavior when a crowd gathers in a specific occasion as described in any of the above embodiments.

[0031] The embodiment of the present application also provides a system, which includes the device for identifying abnormal behavior when a crowd gathers in a specific occasion as described in the above embodiment.

[0032] An embodiment of the present application also provides a computer-readable storage medium, in which a computer program is stored. When the computer program is executed on a computer, the computer executes the method for identifying abnormal behavior when a crowd gathers in a specific occasion as described in any of the above embodiments.

[0033] The embodiment of the present application provides a method for identifying behavioral abnormalities when a crowd gathers in a specific occasion. When a crowd is detected to be gathered in a specific occasion, the crowd density of the crowd is obtained, and then it is determined whether the crowd density reaches a preset crowd density threshold. If so, it is determined that there is a dense crowd. At this time, the degree of correlation of the crowd is obtained, and then the proportion of outsiders in the crowd is obtained, as well as the time factor and the movement trajectory of the crowd. Finally, the crowd density, degree of correlation, proportion of outsiders, time factor and movement trajectory are input into a pre-trained abnormality deep learning model. The abnormality of the crowd is identified through the abnormality deep learning model, so that the current abnormality of the crowd in the specific occasion (such as group conflict, sudden illness, normal gathering, etc.) can be judged to prepare for the corresponding treatment measures to be taken next. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application, and those skilled in the art can obtain other drawings based on these drawings without creative work.

[0035] Figure 1 A flow chart of a method for identifying abnormal behavior when a crowd gathers in a specific situation provided in an embodiment of the present application.

[0036] Figure 2 A schematic diagram of the structure of a device for identifying abnormal behavior when a crowd gathers in a specific occasion provided by an embodiment of the present application. DETAILED DESCRIPTION

[0037] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application.

[0038] The present application embodiment provides a method for identifying abnormal behavior when a crowd gathers in a specific occasion. Figure 1 , Figure 1 A flow chart of a method for identifying abnormal behavior when a crowd gathers in a specific occasion provided by an embodiment of the present application, the method may include the following steps:

[0039] Step 101, when it is detected that a crowd gathers in a specific occasion, the crowd density of the crowd is obtained.

[0040] Among them, the specific occasions may be industrial parks, large gatherings, transportation hubs, commercial centers and other specific occasions, and the specific occasion in the embodiment of the present application is an industrial park.

[0041] In some embodiments, obtaining the crowd density of the crowd includes:

[0042] Obtain the area of ​​the specific occasion;

[0043] Get the total number of people in the specific situation;

[0044] The crowd density is generated based on the area of ​​the region and the total number of people.

[0045] Specifically, the crowd density can be calculated as follows:

[0046] D=N total / A area

[0047] Among them, N total is the total number of people, A area is the area of ​​the region.

[0048] Step 102, determining whether the crowd density reaches a preset crowd density threshold, if so, determining that there is a crowd-dense situation, and obtaining the degree of association of the crowd.

[0049] Among them, those skilled in the art can set the preset crowd density threshold according to actual conditions, and no specific limitation is made here.

[0050] Among them, the reasons for the existence of dense crowds may be group conflicts, sudden illnesses, work-related protests, normal gatherings, etc.

[0051] In some embodiments, the association degree includes familiarity, and obtaining the association degree of the group of people includes:

[0052] Obtain the number of first persons in the group of people who belong to the same company, and obtain the number of second persons in the group of people whose company belongs to the same floor;

[0053] Get the total number of people in the group;

[0054] The familiarity of the group of people is generated based on the first number of people, the second number of people and the total number of people.

[0055] Specifically, familiarity can be calculated as follows:

[0056] F=(N same_company +N same_floor ) / N total

[0057] Among them, N same_company is the number of people belonging to the same company (the first number of people), N same_floor is the number of people on the same floor (the second number of people), N total is the total number of people.

[0058] In some embodiments, before obtaining the first number of people in the group who belong to the same company and obtaining the second number of people in the group whose company belongs to the same floor, the method further includes:

[0059] The basic information of each person in the crowd is obtained, wherein the basic information includes the company to which the person belongs and the floor on which the person's company is located.

[0060] In some embodiments, the association degree includes interaction intensity, and obtaining the association degree of the group of people includes:

[0061] Obtaining facial expressions, voice information, and body movements of each person in the group;

[0062] The facial expression, voice information and body movement are input into a pre-trained deep learning model, and the interaction intensity is generated by the deep learning model.

[0063] Specifically, the interaction intensity can be calculated by analyzing data such as body movements, facial expressions, and voice information between people. The present embodiment uses a deep learning model (recurrent neural network RNN) to predict the interaction intensity:

[0064] I = f (body movements, facial expressions, chat voice)

[0065] Among them, ff is a trained deep learning model.

[0066] Step 103, obtaining the proportion of outsiders in the crowd, as well as obtaining the time factor and the movement trajectory of the crowd.

[0067] Among them, outsiders may bring more uncertainty.

[0068] The time factor can be calculated by the behavior pattern in the historical data. Specifically, the embodiment of the present application has a time-based behavior pattern model g(t), which can predict the normal behavior pattern in a specific time period:

[0069] T = g(t)

[0070] Where t is the current time.

[0071] Among them, the movement trajectory of the crowd is obtained through the Reid feature extraction algorithm.

[0072] In some embodiments, obtaining the proportion of outsiders in the population includes:

[0073] Get the total number of people;

[0074] Get the number of foreigners in the population;

[0075] The foreign personnel ratio is generated based on the number of the foreign personnel and the total number of the population.

[0076] Specifically, the proportion of foreign personnel can be calculated in the following way:

[0077] O=N outsiders / N total

[0078] Among them, N outsiders is the number of outsiders, N total is the total number of people (the total number of people).

[0079] Step 104: input the crowd density, correlation degree, proportion of outsiders, time factor and action trajectory into a pre-trained abnormality deep learning model, and identify the abnormality of the crowd through the abnormality deep learning model.

[0080] Among them, when the degree of association includes familiarity and interaction intensity, the crowd density, familiarity, interaction intensity, proportion of outsiders, time factors and action trajectory are input into a pre-trained anomaly deep learning model, and the abnormality of the crowd is identified through the anomaly deep learning model.

[0081] The abnormality of crowd gathering is calculated by combining crowd density, familiarity, interaction intensity, proportion of outsiders, time factors and movement trajectories. For example, if people are unfamiliar with each other and interact frequently, there may be group conflict; if people are familiar with each other but suddenly show abnormal behavior, it may be a sudden illness, etc.

[0082] Among them, in certain occasions, the behaviors of people gathering may include the following two: ① someone suddenly falls down, and other people around come to watch; ② someone gets injured at work, and then a group of people carry the injured to ask the company boss for compensation. For these two situations, the existing technology cannot determine which one it is, but the technical means of this application can determine it.

[0083] Among them, the embodiment of the present application adopts an abnormality deep learning model to comprehensively consider the above factors and output the final abnormality. Specifically, the embodiment of the present application uses a fully connected neural network (Fully Connected Neural Network, FCNN) to calculate the abnormality. The situation of the fully connected neural network (abnormality deep learning model) is as follows:

[0084] 1. Input layer

[0085] The input layer includes the following features:

[0086] F: familiarity; O: proportion of outsiders; I: interaction intensity; D: crowd density; T: time factor; R: action trajectory. Among them, R = Reid(p), p is each person, and R is each person's trajectory.

[0087] 2. Output layer

[0088] Feature extraction: Get data X = [F, O, I, D, T, R]

[0089] Normalization:

[0090] Each feature is normalized: x′=(x-μ) / σ, where μ is the mean of the feature and σ is the standard deviation of the feature.

[0091] Get the normalized data X = [F′, O′, I′, D′, T′, R′]

[0092] 3. Hidden Layer

[0093] The hidden layer uses multiple fully connected layers, and each layer uses the ReLU activation function:

[0094] h1=ReLU(W1·X+b1)

[0095] h2=ReLU(W2·h1+b2)

[0096] …

[0097] h n =ReLU(W n ·h n-1 +b n )

[0098] 4. Output Layer

[0099] The output layer uses the Sigmoid activation function to output an abnormality value between 0 and 1:

[0100] A=σ(W o ·h n +b o )

[0101] Where σ is the Sigmoid function: σ(x) = 1 / (1+e -x )

[0102] 5. Loss Function

[0103] The embodiment of the present application uses a binary cross-entropy loss function to train the model. Assuming that the label y indicates whether it is an abnormal behavior (0 indicates normal, 1 indicates abnormal), the loss function LL can be expressed as: L = -ylog(A)-(1-y)log(1-A).

[0104] 6. Training Process

[0105] ① Data preparation: Collect a large amount of actual data, including characteristics such as familiarity, proportion of outsiders, interaction intensity, crowd density, time factors, and corresponding labels (whether it is abnormal behavior).

[0106] ②Model construction: Construct the above-mentioned fully connected neural network model.

[0107] ③Model training: Use the gradient descent method (Adam optimizer) to minimize the loss function L and update the model parameters W and b.

[0108] ④ Model validation: Evaluate the model performance through the cross-validation method, and select the optimal model as the abnormality deep learning model of the embodiment of the present application.

[0109] In some embodiments, the abnormality includes normal and abnormal, and after identifying the abnormality of the population, the method further includes:

[0110] If the abnormality is abnormal, corresponding processing measures are triggered based on the abnormal situation.

[0111] Among them, in specific occasions, crowd gathering behaviors can be divided into the following categories:

[0112] ① Group conflict: refers to conflicts such as disputes and fights among a group of people.

[0113] ②Sudden illness: refers to someone in the crowd suddenly falling ill and requiring emergency medical assistance.

[0114] ③Normal gatherings: refers to normal gatherings, meetings and other non-abnormal behaviors.

[0115] ④Others: other normal and abnormal aggregation situations.

[0116] The abnormality in the embodiment of the present application refers to group conflicts, sudden illnesses, and other abnormal gatherings, and the normality refers to normal gatherings or other normal situations. Among them, the existing methods have the following limitations when identifying these different types of behaviors:

[0117] ① Single feature reliance: Most methods rely on a single feature (such as population density, motion trajectory, etc.) and are unable to comprehensively consider multiple factors.

[0118] ② Lack of contextual information: Failure to fully utilize contextual information (such as relationships between people, interaction patterns, etc.) to make more accurate judgments.

[0119] ③ It is difficult to distinguish complex behaviors: Existing methods are difficult to effectively distinguish and identify complex behaviors such as group conflicts and sudden illnesses.

[0120] However, the above limitations no longer exist through the abnormality identification method of the present application.

[0121] Among them, the handling measures include sending alarms, dispatching security personnel, and activating emergency plans.

[0122] This application combines advanced computer vision technology and deep learning algorithms, which can accurately and efficiently identify abnormal behaviors in complex and changing environments, and take corresponding treatment measures in a timely manner to improve the level of public safety.

[0123] The advantages of this application are as follows:

[0124] 1. High-precision abnormal behavior identification

[0125] Multimodal data fusion: By combining data from multiple modalities such as body movements, facial expressions, and voice information, it is possible to capture the behavioral characteristics of the crowd more comprehensively and improve the accuracy of abnormal behavior identification.

[0126] Deep learning model: Using deep learning models for feature extraction and classification can effectively capture complex behavior patterns and improve recognition accuracy.

[0127] 2. Consider multiple factors

[0128] Familiarity: By calculating the familiarity between people (such as whether they belong to the same company or are on the same floor), it is possible to more accurately determine the reasons for crowd gatherings and distinguish between normal gatherings and potential abnormal behaviors.

[0129] Proportion of outsiders: Analyzing the proportion of outsiders can help identify potential safety hazards.

[0130] Interaction intensity: By analyzing the intensity of interaction between people, group conflicts or other abnormal behaviors can be judged more accurately.

[0131] Crowd Density: Considering changes in crowd density helps identify potential risks in crowded situations.

[0132] Time factor: Incorporating the time factor can help us better understand the normal patterns of crowd behavior in different time periods, thereby more accurately identifying abnormal behavior.

[0133] Action trajectory: Considering that different crowd gathering behaviors have different action trajectories, combining action trajectories can more accurately determine the reasons for crowd gathering, thereby distinguishing different gatherings. For example, for group conflicts, the action trajectory of the crowd is reflected as follows: ① Changes in gathering density: In group conflicts, people tend to gather quickly and form high-density areas; ② Movement direction and speed: People in conflict usually have fast and disorderly movements, which is different from the orderly movement in normal gatherings; ③ Interaction pattern: People in conflict may frequently approach and move away from each other, showing intense interaction. For sudden illness, the action trajectory of the crowd is reflected as follows: ① Sudden stop: If someone suddenly stops moving and does not move for a long time, it may be a sign of sudden illness; ② Help-seeking behavior: The sick person may seek help from the people around him, which is manifested as a specific action trajectory, such as slow movement or falling down; ③ Crowd reaction: The people around him may surround the sick person, forming a local high-density area, and may have obvious interactive behaviors (such as calling an ambulance). For normal gatherings, the action trajectory of the crowd is reflected as follows: ① Orderly movement: In normal gatherings, people's movements are usually orderly, such as queuing, walking, etc. ② Low-density areas: The distribution of people in normal gatherings is relatively uniform, and there will be no extreme high-density areas. ③ Stable interaction: Normal interaction patterns are usually stable and predictable, without drastic changes. Therefore, combining the action trajectory can more accurately determine the reasons for the gathering of people, thereby distinguishing different gatherings.

[0134] As can be seen from the above, this application aims to timely discover and deal with potential safety hazards by intelligently analyzing and identifying the behavior patterns of people in specific occasions. This application introduces a multi-weight factor method, comprehensively considers multiple factors such as the relationship between people and interaction patterns, and can more accurately identify and distinguish different types of gathering behaviors, thereby improving the level of public safety management. This method not only overcomes the limitations of existing technologies, but also provides strong technical support for intelligent management in the field of public safety.

[0135] All the above optional technical solutions can be arbitrarily combined to form optional embodiments of the present application, which will not be described one by one here.

[0136] In specific implementation, the present application is not limited by the execution order of the various steps described. If no conflict occurs, some steps can be performed in other orders or simultaneously.

[0137] From the above, it can be seen that the method for identifying behavioral abnormalities when a crowd gathers in a specific occasion provided by the embodiment of the present application, when a crowd is detected to be gathered in a specific occasion, the crowd density of the crowd is obtained, and then it is determined whether the crowd density reaches a preset crowd density threshold. If so, it is determined that there is a dense crowd situation, and the degree of correlation of the crowd is obtained at this time, and then the proportion of outsiders in the crowd is obtained, as well as the time factor and the movement trajectory of the crowd. Finally, the crowd density, degree of correlation, proportion of outsiders, time factor and movement trajectory are input into a pre-trained abnormality deep learning model, and the abnormality of the crowd is identified through the abnormality deep learning model, so that the current abnormality of the crowd in a specific occasion (such as group conflict, sudden illness, normal gathering, etc.) can be judged to prepare for the corresponding treatment measures to be taken next.

[0138] The present application also provides a device for identifying abnormal behavior when a crowd gathers in a specific occasion. Figure 2 , Figure 2 A schematic diagram of a structure of a device for identifying abnormal behavior when a crowd gathers in a specific occasion provided by an embodiment of the present application. The device 30 for identifying abnormal behavior when a crowd gathers in a specific occasion includes a memory 120, one or more processors 180, and one or more applications, wherein the one or more applications are stored in the memory 120 and configured to be executed by the processor 180; the memory 120 can be used to store applications and data. The application stored in the memory 120 contains executable code. The application can be composed of various functional modules. The processor 180 executes various functional applications and data processing by running the application stored in the memory 120. In addition, the memory 120 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other volatile solid-state storage device. Accordingly, the memory 120 may also include a memory controller to provide the processor 180 with access to the memory 120.

[0139] The processor 180 is the control center of the device, and uses various interfaces and lines to connect various parts of the entire terminal, and executes various functions of the device and processes data by running or executing applications stored in the memory 120, and calling data stored in the memory 120, so as to monitor the device as a whole. Optionally, the processor 180 may include one or more processing cores; preferably, the processor 180 may integrate an application processor and a modem processor, wherein the application processor mainly processes the operating system, user interface, and application programs.

[0140] Specifically in this embodiment, a computer program is stored in the memory 120, and the processor 180 executes the method for identifying abnormal behavior when a crowd gathers in a specific occasion as described in any of the above embodiments by calling the computer program stored in the memory 120.

[0141] An embodiment of the present application also provides a system, which includes the above-mentioned device for identifying abnormal behavior when a crowd gathers in a specific occasion.

[0142] An embodiment of the present application further provides a storage medium, in which a computer program is stored. When the computer program is executed on a computer, the computer executes the method for identifying abnormal behavior when a crowd gathers in a specific occasion as described in any of the above embodiments.

[0143] It should be noted that, for the method for identifying abnormal behavior when a crowd gathers in a specific occasion described in the present application, ordinary testers in the field can understand that the whole or part of the process of implementing the method for identifying abnormal behavior when a crowd gathers in a specific occasion described in the embodiment of the present application can be completed by controlling the relevant hardware through a computer program, and the computer program can be stored in a computer-readable storage medium, such as a memory of a system, and executed by at least one processor in the system, and the execution process may include the process of the embodiment of the method for identifying abnormal behavior when a crowd gathers in a specific occasion. Among them, the storage medium can be a disk, an optical disk, a read-only memory (ROM, Read Only Memory), a random access memory (RAM, Random Access Memory), etc.

[0144] The above is a detailed introduction to the method, device, storage medium and system for identifying abnormal behavior when a crowd gathers in a specific occasion provided by the embodiments of the present application. This article uses specific examples to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method and core ideas of the present application; at the same time, for those skilled in the art, according to the ideas of the present application, there will be changes in the specific implementation methods and application scopes. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A method for identifying abnormal behavior of a crowd in a specific occasion, characterized in that: The method comprises: When a crowd is detected to be gathered in a specific occasion, obtaining the crowd density of the crowd; Determine whether the crowd density reaches a preset crowd density threshold, and if so, determine that there is a crowd density condition, and obtain the correlation degree of the crowd; Obtaining the proportion of outsiders in the population, as well as obtaining the time factor and the movement trajectory of the population; The crowd density, correlation degree, proportion of outsiders, time factor and action trajectory are input into a pre-trained abnormality deep learning model, and the abnormality of the crowd is identified through the abnormality deep learning model.

2. The method for identifying abnormal behavior of a crowd gathering in a specific occasion as claimed in claim 1, characterized in that: The association degree includes familiarity, and obtaining the association degree of the group of people includes: Obtain the number of first persons in the group of people who belong to the same company, and obtain the number of second persons in the group of people whose company belongs to the same floor; Get the total number of people in the group; The familiarity level is generated based on the first number of people, the second number of people, and the total number of people.

3. The method for identifying abnormal behavior of a crowd gathering in a specific occasion as claimed in claim 2, characterized in that: Before obtaining the number of first persons in the group who belong to the same company and obtaining the number of second persons in the group who belong to the same company and the same floor, the method further includes: The basic information of each person in the crowd is obtained, wherein the basic information includes the company to which the person belongs and the floor on which the person's company is located.

4. The method for identifying abnormal behavior of a crowd gathering in a specific occasion as claimed in claim 1, characterized in that: The degree of association includes interaction intensity, and obtaining the degree of association of the group of people includes: Obtaining facial expressions, voice information, and body movements of each person in the group; The facial expression, voice information and body movement are input into a pre-trained deep learning model, and the interaction intensity is generated by the deep learning model.

5. The method for identifying abnormal behavior of a crowd gathering in a specific occasion as claimed in claim 1, characterized in that: The abnormality includes normal and abnormal. After identifying the abnormality of the group, the method further includes: If the abnormality is abnormal, corresponding processing measures are triggered based on the abnormal situation.

6. The method for identifying abnormal behavior of a crowd gathering in a specific occasion as claimed in claim 1, characterized in that: The obtaining the proportion of outsiders in the population includes: Get the total number of people; Get the number of foreigners in the population; The foreign personnel ratio is generated based on the number of the foreign personnel and the total number of the population.

7. The method for identifying abnormal behavior of a crowd gathering in a specific occasion as claimed in claim 1, characterized in that: The obtaining of the crowd density of the crowd includes: Obtain the area of ​​the specific occasion; Get the total number of people in the specific situation; The crowd density is generated based on the area of ​​the region and the total number of people.

8. A device for identifying abnormal behavior of a crowd in a specific occasion, characterized in that: The device includes a processor and a memory, wherein a computer program is stored in the memory, and the processor is used to execute the method for identifying abnormal behavior when a crowd gathers in a specific occasion as described in any one of claims 1 to 7 by calling the computer program stored in the memory.

9. A system, characterized in that: The system includes the device for identifying abnormal behavior when a crowd gathers in a specific occasion as described in claim 8.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed on a computer, the computer executes the method for identifying abnormal behavior when a crowd gathers in a specific occasion as described in any one of claims 1 to 7.

Images