CN112149457A - People flow statistical method, device, server and computer readable storage medium - Google Patents

Abstract

The present invention provides a method, device, server and computer-readable storage medium for people flow statistics. The method includes: collecting a first group of frame images and a second group of frame images, the first group of frame images are from a first group of cameras set in the deployment area, and the second group of frame images are from a second group of camera devices set in the deployment area , and the shooting angle of view of the first group of cameras faces the direction of the first flow of people, and the shooting angle of view of the second group of camera devices faces the direction of the second flow of people; respectively perform avatar detection on the first group of frame images and the second group of frame images; structured storage The avatar information detected based on the first set of frame images and the avatar information detected based on the second set of frame images; the number of people flowing in the first crowd direction is calculated according to the avatar information of the first set of frame images and the avatar information of the second set of frame images and the number of people in the second direction of flow. Based on the present invention, it is relatively simple to realize the statistics of the flow of people in each direction of the flow of people in the control area.

Description

Method, apparatus, server, and computer-readable storage medium for people flow statistics

technical field

The present invention relates to the field of computer technology, and in particular, to a method, device, server and computer-readable storage medium for people flow statistics.

Background technique

Judging and preventing crowded and stampede events in high-traffic areas and places is a key concern of government departments. To prevent and deploy in advance is a problem that government departments are currently working on. How to accurately determine the real-time flow of people in key areas and places , There are also some methods and experiences in the direction of people flow. The main statistical methods of people flow are: manual statistics and bayonet flow statistics. The advantages and disadvantages of these two methods are analyzed as follows.

1) Manual statistics

Manual statistics is a traditional way of counting people flow. Statisticians are deployed through entrances and exits to count the number of people entering and leaving in real time, and then report and calculate the number of people in the current area. This method requires a lot of manpower and requires the attention of the statisticians. It is highly concentrated and prone to errors. It can only count the changes in the number of people, but cannot obtain the flow of people in a specific area. It is difficult to implement in places with many exits and entrances, and it is time-consuming and laborious to cooperate with facilities such as enclosures.

2) Bayonet traffic statistics

The manual counting of people is replaced by artificial intelligence equipment. The main entrances and exits are equipped with bayonet equipment for people counting. The real-time number of people in the current defense range is calculated through statistical data, and the flow of people is diverted according to the statistical data. The advantage of this method is that it can count the number of people more accurately, but it is impossible to make accurate judgments on the information such as the specific flow of people in the control area, and it still needs to be identified manually.

It can be seen that the existing statistical methods of people flow cannot achieve the desired effect.

SUMMARY OF THE INVENTION

In view of this, the present invention provides a method, an apparatus, a server and a computer-readable storage medium for people flow statistics, so as to solve the problems existing in the existing people flow statistics in the control area.

According to a first aspect of the present invention, there is provided a method for counting people flow, comprising the following steps:

Collect a first group of frame images and a second group of frame images, the first group of frame images are from the first group of cameras set in the deployment area, and the second group of frame images are from the first group of frame images set in the deployment area. Two sets of camera devices, and the shooting angle of view of the first group of camera devices faces the direction of the first flow of people, and the shooting angle of view of the second group of camera devices faces the direction of the second flow of people;

Perform avatar detection on the first group of frame images and the second group of frame images respectively;

Structured storage of avatar information detected based on the first set of frame images and avatar information detected based on the second set of frame images; and

According to the avatar information of the first group of frame images and the avatar information of the second group of frame images, the number of people flow in the first direction of people flow and the number of people flow in the second group of people flow direction are calculated.

In an optional embodiment, determining whether there is an overlapping area between the shooting range of the first group of camera devices and the shooting range of the second group of camera devices;

If there is an overlapping area between the shooting range of the first group of camera devices and the shooting range of the second group of camera devices, calculate the area ratio of the overlapping area to the control area, and correct the first group according to the area ratio. The number of people in the flow direction of one person and the number of people flow in the second flow direction.

In an optional embodiment, the method further includes: displaying the number of people flow in the first direction of people flow and the number of people flow in the second direction of people flow.

In an optional embodiment, the method further includes: comparing the number of people in the first direction of people flow and the number of people in the second direction of people flow with a set threshold respectively;

If the number of people flow in the first direction of people flow exceeds the set threshold, and/or the number of people flow in the second direction of people flow exceeds the set threshold, a corresponding warning is performed.

In an optional embodiment, performing avatar detection and recognition on the first group of frame images and the second group of frame images respectively includes:

Adopt the YOLO model to output the human head images of the first group of frame images and the second group of frame images respectively; and

The MTCNN model is used to perform aggregation processing on the human head images, so as to distinguish the avatar information of the first human flow direction and the avatar information of the second human flow direction.

In an optional embodiment, the first group of frame images and the second group of frame images are respectively frame images collected at the same time point.

In an optional embodiment, the first people flow direction and the second people flow direction are opposite directions.

According to a second aspect of the present invention, there is provided a device for counting people flow in a control area, including:

The image acquisition module is used to collect a first group of frame images and a second group of frame images, the first group of frame images are from the first group of cameras set in the control area, and the second group of frame images are from the A second group of camera devices arranged in the deployment control area, and the shooting angle of the first group of camera devices faces the direction of the first flow of people, and the shooting angle of the second group of camera devices faces the direction of the second flow of people;

a detection and recognition module, which is used to detect and recognize the avatars of the first group of frame images and the second group of frame images respectively;

A structured storage module for structured storage of the avatar information detected based on the first group of frame images and the avatar information detected based on the second group of frame images; and

A data statistics module, configured to count the number of people flow in the first direction of people flow and the number of people flow in the second direction of people flow based on the avatar information of the first group of frame images and the second group of frame images.

In an optional embodiment, it further includes: a data correction module, configured to determine whether there is an overlapping area between the shooting range of the first group of camera devices and the shooting range of the second group of camera devices, if the first group of camera devices has an overlapping area There is an overlapping area between the shooting range of the group of camera devices and the shooting range of the second group of camera devices, calculate the area ratio of the overlapping area occupying the control area, and correct the flow of people in the first flow direction according to the area ratio The number and the number of people flow in the second flow direction.

In an optional embodiment, the method further includes: a data display module, configured to display the number of people flow in the first direction of people flow and the number of people flow in the second direction of people flow.

In an optional embodiment, it includes: an early warning module, used for an early warning module, for comparing the number of people in the first direction of people flow and the number of people in the second direction of people flow with a set threshold respectively; if If the number of people flow in the first direction of people flow exceeds the set threshold, and/or the number of people flow in the second direction of people flow exceeds the set threshold, a corresponding warning is issued.

In an optional embodiment, the detection and identification module includes:

The first detection unit is used to output the human head images of the first group of frame images and the second group of frame images respectively using the YOLO model;

The image alignment unit is used to process the head image by using the MTCNN model, so as to distinguish the attributes of the avatar, and the attributes of the avatar include the front face, the side face and the back of the head.

According to a third aspect of the present invention, a server is provided, comprising:

processor;

memory for storing processor-executable instructions;

Wherein, the processor is configured to perform any of the methods described above.

According to a fourth aspect of the present invention, a computer-readable storage medium is provided, where the computer-readable storage medium stores computer instructions, and when the computer instructions are executed, implements any one of the methods described above.

The present invention has the following effects: by arranging camera devices in different people flow directions, collecting frame images, and performing statistics on the number of people flow in each direction according to the frame images, it is relatively simple to realize the statistics of people flow in each direction in the entire control area, So as to solve the current situation that the security department needs to deploy a large number of manpower for flow reporting and dredging.

Description of drawings

The above and other objects, features and advantages of the present invention will become more apparent from the description of embodiments of the present invention with reference to the following drawings, in which:

FIG. 1 is a schematic diagram of an exemplary deployment area;

Fig. 2 is the connection schematic diagram of camera device and server;

FIG. 3 is a flowchart of a method for counting people flow in a control area provided according to an embodiment of the present invention;

FIG. 4 is a structural diagram of an apparatus for counting people flow in a control area provided according to an embodiment of the present invention;

FIG. 5 is a structural diagram of a server for performing statistics on people flow in a control area according to an embodiment of the present invention.

Detailed ways

The present invention is described below based on examples, but the present invention is not limited to these examples only. In the following detailed description of the invention, some specific details are described in detail. The present invention can be fully understood by those skilled in the art without the description of these detailed parts. In order to avoid obscuring the essence of the present invention, well-known methods, procedures and processes are not described in detail. Additionally, the drawings are not necessarily to scale.

The flowcharts and block diagrams in the accompanying drawings illustrate the possible architecture, functions, and operations of the systems, methods, and devices of the embodiments of the present invention, and the blocks on the flowcharts and block diagrams may represent a module, a program segment, or just a segment. Code, the modules, program segments and codes are all executable instructions for implementing specified logical functions. It should also be noted that the executable instructions implementing the specified logical functions may be recombined to create new modules and program segments. Therefore, the blocks and the block sequence of the drawings are only used to better illustrate the process and steps of the embodiment, and should not be taken as a limitation on the invention itself.

FIG. 1 is a schematic diagram of an exemplary deployment area. As shown in FIG. 1 , the deployment area 100 is an S-shaped channel, and columns 11-14 for arranging camera devices are arranged at intervals in the channel, and beams (not shown in the figure) may be arranged on the columns 11-14. In this way, the camera device can be set on the beam, and of course, the camera device 1-8 can also be directly set on the column 11-14, wherein the shooting angles of the camera devices 1, 3, 5 and 7 are directed towards the first flow of people. The shooting angles of the devices 2, 4, 6 and 8 are directed towards the second flow of people. Therefore, the camera devices are divided into a first group of camera devices and a second group of camera devices according to different directions of the shooting angle of view. As shown in FIG. 2 , the imaging devices 1 , 3 , 5 and 7 are the first group of imaging devices, and the imaging devices 2 , 4 , 6 and 8 are the second group of imaging devices. The camera devices 1 to 8 are respectively connected to the server 201 in a wired or wireless manner, and upload the video captured in real time to the server 201 .

We know that the shooting angle determines the shooting range of the camera device. Continuing to refer to FIG. 1 , 30 shows the shooting range of the camera device 5 in the first group of camera devices, and the area 20 shows the shooting range of the camera device 4 in the second group of camera devices. As can be seen from the figure, there is no overlapping area between the area 20 and the area 30, and basically all individuals between the two cameras can be captured in the video. Wherein, the individual traveling in the first direction of people flow is an image of the back of the head in the frame image captured by the camera 5 , and the individual traveling in the second direction of people flow is a front face or side face image in the frame image captured by the camera device 5 . The opposite is true for the camera device 4 . As can be seen from the figure, adding the number of images of the back of the head in the frame image of the camera device 5 plus the number of frontal face images in the camera device 4 is exactly equal to the number of people flow in the first direction of people flow between the camera devices 4 and 5, However, adding the number of frontal (side face) images in the frame image of the camera 5 to the number of the back of the head images in the camera 4 is exactly equal to the number of people in the second direction of people flow between the cameras 4 and 5 . Of course this is the ideal situation. In actual situations, the spacing distance of the columns and the shooting angle of the camera device can be adjusted according to actual needs to achieve the ideal situation.

Combining with practice, it can be seen that the control area 100 is generally a place with a large flow of people such as stations, airports, shopping plazas, tourist attractions, etc. In these places, it is very important to detect, judge and predict the flow direction and number of people. Only by accurately detecting, judging, and predicting the flow direction and the number of people, can the anti-crowding stampede be prevented in advance and provide a reliable basis for the flow restriction and diversion of people. Therefore, in the following embodiments, the present invention provides a method for counting people flow in a control area.

FIG. 3 is a flowchart of a method for counting people flow in a deployment area according to an embodiment of the present invention. The method specifically includes the following steps.

In step S301, a first group of frame images and a second group of frame images are collected.

Wherein, the first group of frame images are from the first group of cameras set in the deployment area, the second group of frame images are from the second group of cameras set in the deployment area, and the first group of cameras sets the camera's Orientation, the second group of camera devices set the orientation of the camera devices according to the second flow direction of people. The first and second pedestrian flow directions are related to the pedestrian flow passages designed in the control area. For example, the first and second pedestrian flow directions in Figure 1 are opposite directions, that is, a 180-degree angle between the two. It should be pointed out that although in FIG. 1 , the first group of camera devices and the second group of camera devices each include four camera devices, however, the present invention is not limited to this, for example, the first group of camera devices and the second group of camera devices Both can contain only one camera.

With reference to FIG. 2, after the server receives the video data captured by the first group of cameras and the second group of cameras in real time, it extracts frame images of multiple key frames from the video data captured by the first group of cameras as the first group of frame images. , extracting frame images of a plurality of key frames from the video data captured by the second group of cameras as the second group of frame images. The video data may be in any media format, for example, the streaming media format established by the Real Time Streaming Protocol (RTSP), and the extracted frame images may be in any picture format.

In step S302, avatar detection is performed on the first group of frame images and the second group of frame images respectively.

In this step, the extracted frame images are detected based on the target detection model. The target detection model is a trained neural network model, and the head images included in the first group of frame images and the second group of frame images are marked respectively through the target detection model to obtain head portrait information. The head image here includes a front face image, a side face image, and a back head image of the person.

In step S303, the avatar information detected based on the first group of frame images and the avatar information detected based on the second group of frame images are stored in a structured manner.

In this step, a database and a data structure for storing avatar information are determined, and the avatar information detected based on the first group of frame images and the avatar information detected based on the second group of frame images are integrated and stored accordingly.

In an optional embodiment, a relational database is used to store avatar information, and the table structure of the corresponding database table includes a plurality of fields such as avatar ID, human head attribute, device number, frame number, etc. The value of the field is stored in the database table. The head attribute is, for example, one of three options: frontal face, profiled face, and back of the head. The equipment number is the unique number of the camera, and it is determined that the camera belongs to the first group or the second camera according to the equipment number. The frame number is used to uniquely identify a frame image, so it can be represented by the time stamp when the frame image was collected. If an avatar is expected to be used in a later step, the avatar image can be stored in the database table in an image format (eg, image).

In another optional embodiment, a document database is used to store avatar information, each document data includes an image and metadata information, the image is, for example, an image of the front face, the side face and the back of the head, and the metadata information is an abstract corresponding to the image information, the summary information includes, for example, an avatar ID, head attributes, pixel characteristics, and the like.

In step S304, according to the avatar information of the first group of frame images and the avatar information of the second group of frame images, the number of people flow in the first direction of people flow and the number of people flow in the second group of people flow direction are calculated.

In this step, statistical calculation is performed based on the avatar information to obtain the number of people in the first flow direction and the number of people in the second flow direction. For example, the number of people flow can be expressed by the number of individuals within a square kilometer. Among them, the formula for statistical calculation is related to the deployment method of the camera device and the shooting angle of view. Take the deployment method of the camera device mentioned in FIG. 1 as an example. According to the first set of frame images, for example, it is 2000 (front face + side face) + 500 back of the head, and according to the second set of frame images, for example, it is 400 (front face + side face) + 1000 back of the head, then the number of people in the first flow direction It is equal to 2000 of the first set of frame images (front face + side face) + 1000 of the back of the head of the second set of frame images, and the number of people in the second flow direction is equal to 400 of the second set of frame images (front face + side face) + first 500 back heads of the framed images, and then the number of people flow can be calculated according to the number of people flow.

Based on this embodiment, it is relatively simple to realize the statistics of the flow of people in each direction of the flow of people in the entire control area, so as to solve the current situation that the security department needs to deploy a large number of manpower for flow direction reporting and dredging. Understand the current flow of people in various directions in the entire control area and make accurate flow control, and based on this, conduct rapid and unified dredging and deployment to improve the overall response efficiency.

It can be understood that the present invention is limited by the deployment manner of the camera device and the set shooting angle of view. Although the camera device can be manually adjusted to achieve the ideal situation, there may still be duplicate recordings. for example. For example, if a person's front face is included in one frame image and the person's profile face is included in another frame image, the person is counted twice. For another example, if a person's head image is not included in all the current frame images, the person is omitted from the statistics. To this end, an optional way is to integrate the avatar information, integrate the avatar information representing the same person into a record, and then make statistics according to the integrated avatar information; another optional method selects avatars at multiple time points information, and take the average as the statistical result. In addition, in order not to omit the avatar information of the individual, an optional way is to slightly overlap the shooting ranges between the two opposite camera devices, that is, the area 20 and the area 30 in FIG. 1 have overlapping areas. In order to ensure the accuracy of the final counted number of people, the number of people in the first direction of people flow and the number of people in the second direction of people flow obtained in the above embodiment can be corrected according to the size of the overlapping area. Specifically, first count whether there is an overlapping area in the shooting ranges of the first group of photographed images and the second group of photographed images. If there is an overlapping area, calculate the area ratio of the overlapping area to the control area, and then correct the first flow direction according to the area ratio. If there is no overlapping area, it is not necessary to correct the number of people in the first and second directions of human flow obtained in step S304.

In an optional embodiment, in step S302, the YOLO model is used to mark the head images included in the first group of frame images and the second group of frame images, and then the MTCNN model is used to align the head images, and the attributes of the head are output. . The YOLO model is a trained object detection model used to mark the head position (including the front face, side face and back of the head) in the image. The MTCNN model is a trained alignment network model, which is used to mark the position of the facial features in the head image, and then judge that the head is a frontal face image, a side face image and an image of the back of the head. Of course, the implementation of the present invention does not depend on the two neural network models, and other algorithms or models can also be used to implement the present invention.

In the above embodiments, two people flow directions are used as examples for description. It should be understood at the time that the present invention can also be extended to, for example, deployment areas in four directions of flow of people. For example, in the four directions of east, south, west, and north, according to the results, it can be concluded that 20% of the total number of people in the current area go east, 30% go west, 30% go south, and 20% go north, and then security The department can carry out accurate early warning and control of the flow of people at the entrance and exit according to the actual flow of pedestrians. For example, for the control area, set a threshold for the number of people flow. When the number of people in a certain direction or several directions exceeds the set threshold for the number of people, a corresponding warning will be issued. If the number of people in any direction does not exceed the number of people. The threshold, No warning is given. The number of people flow data can be presented through the display screen. Correspondingly, the warning information can also be presented through the display screen.

FIG. 4 is a structural diagram of an apparatus for counting people flow in a deployment area according to an embodiment of the present invention. The device, for example, runs on the server shown in Figure 2, and specifically includes the following modules.

The image acquisition module 401 is used to collect a first group of frame images and a second group of frame images, the first group of frame images is from the first group of cameras set in the deployment area, and the second group of frame images is from the camera set in the deployment area. The second group of camera devices has a shooting angle of view facing the first crowd flow direction, and the shooting angle of the second group of camera devices faces the second crowd flow direction.

The detection and identification module 402 is configured to perform avatar detection on the first group of frame images and the second group of frame images respectively.

The structured storage module 403 is used for structured storage of the avatar information detected based on the first group of frame images and the avatar information detected based on the second group of frame images.

The data statistics module 404 is configured to count the number of people flow in the first direction of people flow and the number of people flow in the second direction of people flow based on the avatar information of the first group of frame images and the second group of frame images.

Based on this device, it is relatively simple to realize the flow statistics of each flow direction in the entire control area, so as to solve the current situation that the security department needs to deploy a large number of manpower for flow direction reporting and dredging, and the security department can accurately understand the flow statistics results. Up to now, the flow of people in each direction of the entire control area is controlled and accurate flow control is made, and based on this, rapid and unified dredging and deployment are carried out to improve the overall response efficiency.

In an optional embodiment, the above-mentioned apparatus further includes: a data correction module, configured to determine whether to determine whether to correct the data according to whether there is an overlapping area between the shooting range of the first group of cameras and the shooting range of the second group of cameras. The number of people in the first direction of people flow and the number of people in the second direction of people flow are corrected.

In an optional embodiment, the above-mentioned apparatus further includes: a data display module, configured to display the number of people in the first direction of people flow and the number of people in the second direction of people flow.

In an optional embodiment, the above-mentioned device further includes: an early warning module, configured to perform an early warning according to the number of people in the first direction of people flow and the number of people in the second direction of people flow.

In an optional embodiment, the detection and identification module includes:

The first detection unit is used to output the human head images of the first group of frame images and the second group of frame images respectively using the YOLO model;

The image alignment unit is used to process the head image by using the MTCNN model, so as to distinguish the attributes of the avatar, and the attributes of the avatar include the front face, the side face and the back of the head.

In an optional embodiment, the first group of frame images and the second group of frame images are respectively multiple frame images collected at the same time interval within the same set period.

In an optional embodiment, the first people flow direction and the second people flow direction are opposite directions.

It should be understood that the embodiments of the apparatus of the present invention correspond to the embodiments of the method. Therefore, in the embodiments of the apparatus, some repetitive contents are not described in detail.

FIG. 5 is a structural diagram of a server according to an embodiment of the present invention. Referring to FIG. 5 , the server device 50 includes at least one processor 510 , a memory 520 , an input device 530 and an output device 540 connected through a bus. The memory 520 includes read-only memory (ROM) and random access memory (RAM). Various computer instructions and data required to perform system functions are stored in the memory 520. The processor 510 reads various computer instructions from the memory 520 to execute. Various appropriate actions and handling. The input and output devices include an input section 5 such as a keyboard, a mouse, etc.; an output section such as a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section including a hard disk, etc.; The communication part of the network interface card. The memory also stores specific computer instructions, the processor 510 reads the specific computer instructions from the memory 520, and executes the instructions to complete the following operations: collecting a first group of frame images and a second group of frame images, the first group of frame images The images are from a first group of cameras set in the deployment area, the second group of frame images are from a second group of cameras set in the deployment area, and the shooting angles of the first group of cameras face the The first crowd flow direction, the shooting angle of the second group of cameras faces the second crowd flow direction; the avatar detection is performed on the first group of frame images and the second group of frame images respectively; the structured storage is based on the The avatar information detected by the first group of frame images and the avatar information detected based on the second group of frame images; The number of people flow in the first direction of people flow and the number of people flow in the second direction of people flow.

Correspondingly, an embodiment of the present invention provides a computer-readable storage medium, where computer instructions are stored in the computer-readable storage medium, and when the computer instructions are executed, the operations specified by the above data processing method are implemented.

The flowcharts and block diagrams in the accompanying drawings illustrate the possible architecture, functions, and operations of the systems, methods, and devices of the embodiments of the present invention, and the blocks on the flowcharts and block diagrams may represent a module, a program segment, or just a segment. Code, the modules, program segments and codes are all executable instructions for implementing specified logical functions. It should also be noted that the executable instructions implementing the specified logical functions may be recombined to create new modules and program segments. Therefore, the blocks and the block sequence of the drawings are only used to better illustrate the process and steps of the embodiment, and should not be taken as a limitation on the invention itself.

Each module or unit of the system can be implemented by hardware, firmware or software. The software includes, for example, coding programs formed in various programming languages such as JAVA, C/C++/C#, and SQL. Although the steps and sequence of the steps in the embodiments of the present invention are given in the method and the method illustration, the executable instructions for implementing the specified logical functions of the steps may be recombined to generate new steps. The sequence of the steps should not be limited only to the sequence of steps in the method and the method legend, and can be adjusted at any time according to functional requirements. For example, some of the steps are performed in parallel or in reverse order.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims (14)

1. a method for people flow statistics, is characterized in that, comprises the following steps: A first group of frame images and a second group of frame images are collected, the first group of frame images are from the first group of camera devices, the second group of frame images are from the second group of camera devices, and the first group of frame images are from the second group of camera devices. The shooting angle of view faces the first crowd flow direction, and the shooting angle of view of the second group of cameras faces the second crowd flow direction; Perform avatar detection on the first group of frame images and the second group of frame images respectively; Structured storage of avatar information detected based on the first set of frame images and avatar information detected based on the second set of frame images; and According to the avatar information of the first group of frame images and the avatar information of the second group of frame images, the number of people flow in the first direction of people flow and the number of people flow in the second group of people flow direction are calculated. 2. The method of claim 1, further comprising: judging whether there is an overlapping area between the shooting range of the first group of imaging devices and the shooting range of the second group of imaging devices; If there is an overlapping area between the shooting range of the first group of camera devices and the shooting range of the second group of camera devices, calculate the area ratio of the overlapping area to the control area, and correct the first group according to the area ratio. The number of people in the flow direction of one person and the number of people flow in the second flow direction. 3 . The method according to claim 1 , further comprising: displaying the number of people in the first direction and the number of people in the second direction. 4 . 4. The method of claim 1, further comprising: Comparing the number of people in the first direction of people's flow and the number of people's flow in the second direction of people's flow with a set threshold respectively; If the number of people flow in the first direction of people flow exceeds the set threshold, and/or the number of people flow in the second direction of people flow exceeds the set threshold, a corresponding warning is performed. 5. The method according to claim 1, wherein performing head portrait detection and recognition on the first group of frame images and the second group of frame images respectively comprises: Adopt the YOLO model to output the human head images of the first group of frame images and the second group of frame images respectively; and The MTCNN model is used to process the human head image, so as to distinguish the attributes of the avatar, and the attributes of the avatar include the front face, the side face and the back of the head. 6 . The method according to claim 1 , wherein the first group of frame images and the second group of frame images are respectively frame images collected at the same time point. 7 . 7. The method according to any one of claims 1 to 6, wherein the first direction of the flow of people and the second direction of the flow of people are opposite directions, and the avatar information comprises the avatar information of the front face, the side face avatar information and the avatar information of the back of the head. 8. A device for people flow statistics, comprising: an image acquisition module, configured to acquire a first group of frame images and a second group of frame images, the first group of frame images are from a first group of cameras, the second group of frame images are from a second group of cameras, and the The shooting angle of view of the first group of camera devices faces the direction of the first flow of people, and the shooting angle of view of the second group of camera devices faces the direction of the second flow of people; a detection and identification module, configured to perform avatar detection on the first group of frame images and the second group of frame images respectively; A structured storage module for structured storage of the avatar information detected based on the first group of frame images and the avatar information detected based on the second group of frame images; and A data statistics module, configured to count the number of people flow in the first direction of people flow and the number of people flow in the second direction of people flow based on the avatar information of the first group of frame images and the second group of frame images. 9. The apparatus of claim 8, further comprising: A data correction module for judging whether there is an overlapping area between the shooting range of the first group of cameras and the shooting range of the second group of cameras, if the shooting range of the first group of cameras and the second group of cameras There are overlapping areas between the shooting ranges of the camera, the area ratio of the overlapping area to the deployment area is calculated, and the number of people in the first direction of pedestrian flow and the number of people in the second direction of pedestrian flow are corrected according to the area ratio. 10 . The device according to claim 8 , further comprising: a data display module, configured to display the number of people in the first direction of people flow and the number of people in the second direction of people flow. 11 . 11. The device according to claim 8, further comprising: an early warning module for comparing the number of people in the first direction of people flow and the number of people in the second direction of people flow with a set threshold respectively ; If the number of people flow in the first direction of people flow exceeds the set threshold, and/or the number of people flow in the second direction of people flow exceeds the set threshold, a corresponding warning is performed. 12. The device according to claim 8, wherein the detection and identification module comprises: The first detection unit is used to output the human head images of the first group of frame images and the second group of frame images respectively using the YOLO model; The image alignment unit is used to process the head image by using the MTCNN model, so as to distinguish the attributes of the avatar, and the attributes of the avatar include the front face, the side face and the back of the head. 13. A server, comprising: processor; memory for storing processor-executable instructions; wherein the processor is configured to perform the method of any one of claims 1 to 7 above. 14. A computer-readable storage medium, characterized in that the computer-readable storage medium stores computer instructions, which implement the method according to any one of claims 1 to 7 when the computer instructions are executed.

Images