A recognition method, system, computing device, and storage medium based on a dynamic face database

Abstract

The present application relates to the technical field of face recognition, and particularly relates to a recognition method and system based on a dynamic face database, a computing device and a storage medium, comprising a face tracking server configured to acquire face information obtained by a channel camera and a regional camera, and to merge the face information obtained by the channel camera and the regional camera; a central face server configured to store face information left by a user when registering remotely, and to respectively identify face information collected by the face tracking server from the channel camera and the regional camera, and to send the identification results to a channel face server and a regional face server; the present application reduces the number of face databases by using face tracking and face recognition technology, forms a face server with multiple face databases, improves the recognition efficiency by accurately simplifying the face databases, and reduces the false recognition rate of face recognition.

Description

Technical Field

[0001] This invention relates to the field of facial recognition technology, and in particular to a recognition method, system, computing device, and storage medium based on a dynamic facial database. Background Technology

[0002] Currently, many airports have facial recognition systems, used for passenger ID verification, security checks, and more. First, by collecting facial data from passengers who register themselves (via apps, mini-programs, etc.) or on-site, all passenger faces are stored in a large facial database. When a passenger needs to authenticate their face, the data is collected in real-time and then used for facial recognition against the database. As passenger traffic increases, the database expands, and consequently, the probability of misidentification also increases.

[0003] In particular, facial recognition technology is used for security checks, such as the facial recognition system at Hohhot Baita Airport. When passengers pass through the double-door turnstiles, they need to go through a smaller turnstile before entering the restricted area. Only passengers whose faces are successfully identified can enter the restricted area for further security checks.

[0004] After passing through the main turnstile, passengers will enter different security checkpoints randomly. This means that when building the facial recognition database, we can only use a large database. As mentioned above, with the increase in the number of facial recognition database entries, the number of passengers being misidentified also increases.

[0005] To address the above situation, this invention proposes a method for dynamically reducing the passenger database. By reducing the size of the passenger database to be identified, the number of negative samples in the face database is reduced. By utilizing face tracking, a multi-layer recognition method is adopted, from channel to region to the central database, thereby improving the face recognition rate. Summary of the Invention

[0006] To improve face recognition rates, this invention proposes a recognition method, system, computing device, and storage medium based on a dynamic face database. The system includes channel cameras, area cameras, a face tracking server, a channel face server, an area face server, and a central face server, wherein:

[0007] The area to be inspected where the user is located is divided into a first inspection area and a second inspection area. The area camera is installed in the first inspection area and the channel camera is installed in the second inspection area.

[0008] A face tracking server is used to acquire face information from channel cameras and area cameras, and to merge the face information acquired by channel cameras and area cameras.

[0009] The central face server is used to store the face information left by users when registering remotely, and to identify the face information collected by the face tracking server from the channel camera and the area camera, and send the identification results to the channel face server and the area face server respectively.

[0010] Channel face server, used to store the recognition results of channel cameras sent by the central face server;

[0011] The regional face server is used to store the recognition results of regional cameras sent by the central face server.

[0012] Furthermore, the inspection area includes multiple secondary inspection areas, each secondary inspection area includes multiple exit gate channels, and each gate is equipped with a channel camera; each secondary inspection area is equipped with a zone camera.

[0013] Furthermore, the central server only stores the facial information of users who are valid on the same day. Valid users are those who purchased tickets on the same day and whose flights have already departed.

[0014] Furthermore, the area camera captures a set of face images with a face quality greater than a first threshold, and the channel camera captures a single face image with a face quality greater than a second threshold. If a face image captured by the channel camera is within the set of face images captured by the area camera, then it is determined that a certain image in the set of face images captured by the area camera has the highest similarity to a face image captured by a certain channel camera. The information from the set of face images captured by the area camera is then merged with the information from the single face image captured by the channel camera, and then updated in both the channel camera and the area camera to continue tracking the user corresponding to that face information. In this invention, the area camera simultaneously captures multiple different face images, and uploads multiple images that meet the clarity requirements as a set of images to the central face server. That is, in this embodiment, the area camera captures face information of multiple different users within its jurisdiction.

[0015] Preferably, the second threshold is greater than the first threshold.

[0016] This invention also provides a recognition method based on a dynamic face database. Face recognition is performed in a recognition system based on a dynamic face database, specifically including the following steps:

[0017] Users register remotely, during which their facial information is collected and stored in a central facial recognition server.

[0018] The channel camera performs face tracking on users in the second inspection area, and establishes a faceID for face information whose face image quality is greater than the second threshold and sends it to the face tracking server;

[0019] The area camera performs face tracking on users in the first area to be inspected, and establishes a faceID for face information whose face image quality is greater than a first threshold and sends it to the face tracking server;

[0020] The face tracking server includes the face IDs corresponding to the face information captured by the channel camera and the area camera, and determines whether the face IDs can be merged. If they can, the face IDs corresponding to the face information captured by the channel camera and the area camera are merged and updated.

[0021] The central face server compares the facial information captured by the regional cameras with the facial information obtained by the user during registration, and saves the N images in the central face server that are closest to the facial information captured by the regional cameras to the regional face server. These N images are used as the base data of the user in the regional face server. The similarity between the two facial information can be calculated using any existing facial similarity calculation method, and the similarity is normalized to between 0 and 1.

[0022] The central face server compares the facial information captured by the channel cameras with the facial information obtained by the user during registration, and saves the image in the central face server that is closest to the facial information captured by the area cameras to the channel face server.

[0023] The central face server determines whether the face image already exists in the regional face server based on the image stored in the channel face server. If it already exists, it saves the image with the highest similarity among the group of images corresponding to the face ID of that image.

[0024] This invention provides a computer device, the computer device comprising:

[0025] One or more processors;

[0026] Memory, used to store one or more programs;

[0027] When the one or more programs are executed by the one or more processors, the one or more processors implement a recognition method based on a dynamic face database.

[0028] The present invention provides a computer-readable storage medium having a computer program stored thereon, characterized in that the program, when executed by a processor, implements a recognition method based on a dynamic face database.

[0029] This invention reduces the number of face databases by using face tracking and face recognition technology, forming a face server with multiple face databases. By precisely simplifying the face databases, it improves recognition efficiency and reduces the false recognition rate of face recognition. Attached Figure Description

[0030] Figure 1 This is a schematic diagram of a recognition system based on a dynamic face database according to the present invention;

[0031] Figure 2 This invention provides a method for facial recognition based on a dynamic face database. Figure 1 ;

[0032] Figure 3 This invention provides a method for facial recognition based on a dynamic face database. Figure 2 ;

[0033] Figure 4 This is a schematic diagram of the multi-level face recognition process in this invention;

[0034] Figure 5 This is a preferred embodiment for identification in this invention;

[0035] Figure 6 This is a schematic diagram of the network topology for face recognition in this invention. Detailed Implementation

[0036] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0037] This invention proposes a recognition method, system, computing device, and storage medium based on a dynamic face database. The system includes channel cameras, area cameras, a face tracking server, a channel face server, an area face server, and a central face server, wherein:

[0038] The area to be inspected where the user is located is divided into a first inspection area and a second inspection area. The area camera is installed in the first inspection area and the channel camera is installed in the second inspection area.

[0039] A face tracking server is used to acquire face information from channel cameras and area cameras, and to merge the face information acquired by channel cameras and area cameras.

[0040] The central face server is used to store the face information left by users when registering remotely, and to identify the face information collected by the face tracking server from the channel camera and the area camera, and send the identification results to the channel face server and the area face server respectively.

[0041] Channel face server, used to store the recognition results of channel cameras sent by the central face server;

[0042] The regional face server is used to store the recognition results of regional cameras sent by the central face server.

[0043] Passengers register their faces on-site or via app. When they complete check-in and enter the inspection area, the verification channel they enter is random. It's impossible to retrieve passenger facial data from a database using specific information such as passenger identification or flight details, and then accurately send the corresponding facial data to the appropriate front-end device for recognition and verification.

[0044] The passenger facial database will eventually be stored on the backend server. Facial data will be collected by front-end devices and remotely communicated over the network for facial recognition. The backend server will then use the facial information or features sent from the front-end devices to perform corresponding facial recognition.

[0045] When the backend server performs facial recognition, as the facial database grows, the number of security checkpoints increases, and the concurrency rises, the CPU resources consumed by the facial recognition server increase. Simultaneously, a larger facial database also reduces the facial recognition rate. In airports, where rapid passenger clearance and high recognition rates are crucial, fast and accurate recognition are critical performance indicators for facial recognition servers.

[0046] In this embodiment, passengers can register their faces on-site or via an app. Multiple face servers exist in the backend, including a central server that manages the entire face database; regional servers that provide face recognition services for multiple adjacent channels; and terminal servers that perform face recognition for their respective channels. The recognition process based on the dynamic face database includes:

[0047] First, cameras are installed in specific areas. When a passenger enters the identification area through the turnstile, facial tracking technology tracks the area or passage the passenger enters. Then, the central server performs facial recognition to find the passenger's top N faces. These N faces are then sent to the corresponding regional facial recognition server. When a passenger enters a specific area of ​​the passage, the passage camera captures their face and sends it to the central server. The central server finds the top face and sends it to the passage's facial recognition server. The central server also compares this identified face with the top N faces sent to the regional servers. If the face with the highest similarity among the N faces matches this match, the N faces sent to the regional facial recognition server are further narrowed down, retaining only the most similar face.

[0048] Through the above reduction, a channel server can handle approximately 50 people within a certain effective time (from check-in to entering the inspection area), and an area server can handle approximately 150 people if it manages 3 channels, which greatly reduces the number of people in the facial recognition database.

[0049] This embodiment provides a specific implementation method for a recognition system based on a dynamic face database, such as... Figure 1 In this system, each channel has a fixed turnstile, and a camera is installed directly above the channel for facial capture and tracking. Multiple channels are divided into an area, with approximately three cameras controlling each area. During peak hours, the queue may have around 150 people. The facial recognition process includes the following steps:

[0050] (a) Passengers can register their faces on-site or remotely via an app or terminal device.

[0051] (ii) The central facial recognition server is responsible for storing the valid passenger facial recognition database for the current day. Valid passengers refer to those who have purchased tickets for the current flight but whose flight has not yet departed. The passenger information for the central facial recognition server is obtained from other systems within the airport.

[0052] (iii) After entering the inspection area, the passenger will be randomly assigned to either inspection area one or area two.

[0053] (iv) After the area camera captures a passenger, it uses face tracking technology to create a faceId for each captured passenger.

[0054] (V) When the facial quality (facial clarity, contour features) of a captured passenger reaches a certain score, the face recognition interface of the central server is called to store the information of the top N passengers that meet the threshold and inform the tracking service. Since the camera is relatively far away and the monitoring range is large, the facial quality will inevitably be less than ideal. Therefore, the comparison threshold should be slightly lower, for example, around 0.8. This ensures better selection of target faces. Similarly, selecting the top N faces that meet the threshold is also for the purpose of filtering out more target faces and preventing omissions. N should not be too large; a value of around 3 is sufficient, otherwise, the database size will be artificially expanded.

[0055] (vi) Once the target face is selected, the face information is pushed to the regional server so that the regional server can build the face database.

[0056] (vii) The closer a passenger is to the passageway, the greater the chance that the passenger will be captured by the camera above the passageway.

[0057] (viii) When the channel camera captures a face that meets the required quality standard, a face recognition request is triggered, sending a request to the central server to find the best face above the threshold. Because the image clarity is higher at this time, the captured face quality is better, and the recognition threshold is slightly higher than the threshold set when the regional face server built its database, around 0.85. At the same time, the recognized face information is sent to the channel face recognition server.

[0058] (ix) The channel camera will also perform face tracking on each face to avoid repeatedly submitting face recognition requests.

[0059] (x) The channel face recognition server will also send face-related information to the tracking server. The tracking server will use the face information sent by the channel server to further reduce the number of faces in the regional server's face database.

[0060] (xi) In the regional server, each set of N face information is a group of recognition information for the tracked face faceId01. If the tracked face faceid02 in the channel server is the confirmed face information and has the highest similarity among the N face information in this group, then it is considered that faceid01 and faceid02 are tracking the same passenger. In this case, faceids can be merged, and the N face information in the regional server can be reduced to 1 face information.

[0061] (xii) Through the above process, the face database of the central face recognition server is reduced to a regional face database, and then further reduced to a channel-based face recognition database. This layer-by-layer division and reduction ultimately forms a more accurate face database, greatly improving the accuracy and efficiency of face recognition.

[0062] The process of dynamic database creation is as follows: Figures 2-3 As shown, where Figure 2 This refers to the interaction between the channel camera, the area camera, and the face tracking server in this embodiment. Figure 3 This embodiment describes the interaction process between the face tracking server, the channel face server, the regional face service server, the central face server, and the client used for registration. Specifically, it includes the following steps:

[0063] The central face server builds a database based on the user's registration information, i.e., the face database information. In this embodiment, the central face server stores the database information of all face information and is the server with the largest storage capacity in this invention. The next largest is the regional face server, which stores a set of face images as the user's face information. The smallest data volume is that of the channel face server, which only stores the single face information corresponding to the user.

[0064] After the channel camera captures a face, it generates / updates the face tracking faceID and extracts the face quality (any pair of face quality evaluation criteria in the existing technology can be used, which will not be elaborated in this embodiment). If the face quality is greater than the set quality score (this embodiment assumes that the better the face quality, the higher the quality score), the image is sent to the face tracking server; otherwise, it continues to generate / update the face tracking faceID.

[0065] After the area camera captures a face, it generates / updates a face tracking faceID and extracts the face quality (any pair of face quality evaluation criteria in the existing technology can be used, which will not be elaborated in this embodiment). If the face quality is greater than the set quality score (this embodiment assumes that the better the face quality, the higher the quality score), the image is sent to the face tracking server; otherwise, it continues to generate / update the face tracking faceID.

[0066] The face tracking server determines whether to merge the face IDs generated by the area camera and the channel camera based on the recognition results. The face tracking server sends the face ID information obtained from the area camera and the channel camera to the central face server for recognition. This embodiment uses cameras with face recognition and tracking technology to recognize and track faces. The face information acquired by the camera is labeled, that is, each face information is labeled with a face ID. Labeling can improve the recognition accuracy and can also determine the number of users in the area based on the number of face IDs.

[0067] The central face server performs face recognition based on the faceID information captured by the regional cameras, and sends the recognition results to the regional face server. The regional face server receives the Top N face information (i.e., the N face information most similar to the face information to be identified found by the central face server from the registration information), forms a flag as faceID, and uses it to build a database. In this embodiment, the central face server matches N face information in the registration information according to the face data collected by the regional cameras, treats the N face information as a group of faces, and stores this group of faces in the regional face server with faceID as the representation.

[0068] The central face server performs face recognition based on the FaceID information captured by the channel cameras. Each time a channel camera identifies and records a face, it transmits this information to the central face server. The central face server then finds the face with the highest similarity from the registration information based on the received face information, using this as the recognition result corresponding to the face information captured by the channel camera. This recognition result is then sent to both the channel face server and the regional face server. The channel face server builds a database based on the top-ranked face information received (i.e., the face in the registration information that is most similar to the face to be identified). The regional face server checks whether a face database already exists based on the top-ranked face information received. If so, it obtains a set of face data corresponding to the FaceID in that face database and determines the face with the highest similarity among the face data. If the similarity is found, it deletes the face database with that FaceID as its flag and only saves the face with the highest similarity.

[0069] The process of a user entering the inspection area for identification includes:

[0070] When the passenger channel turnstiles are ready to enter the restricted area;

[0071] The turnstile captures the passenger's facial information on-site, finds the best facial image, and sends it to the backend server for recognition;

[0072] After receiving a request from the frontend, the backend server first sends it to the channel server for identification. If successful, it returns a result directly. If it fails, it continues to send the request to the regional server. If successful, it returns a result directly; otherwise, it sends it to the central server. If successful, it returns a result directly; otherwise, it returns a failure result.

[0073] Through the above layers of recognition, a definite result is finally given to the front-end device. In this embodiment, the multi-layer face server will handle most of the requests, and most passengers can be successfully identified in this face database. A few requests will penetrate to the regional server, and even fewer requests will be transmitted to the central server.

[0074] For multi-layered recognition structures, when setting thresholds, the recognition threshold of the channel server should be less than that of the regional server, which in turn should be less than that of the central server. The higher the layer, the more faces there are in the database, and the more interference there will be. Therefore, the threshold for face recognition should be higher to ensure the accuracy of recognition.

[0075] Timing reference of the recognition process Figures 4-5 Specifically, it includes the following steps:

[0076] The user accesses the front-end device (in this embodiment, the front-end device is...) Figure 1 Imported turnstiles allow passage after capturing an image that meets quality requirements, then proceed to the next level of recognition. This involves capturing facial images, extracting facial features, sending facial data from the channel's camera to the channel's facial recognition server, which performs facial recognition. If successful, the data is sent to the front-end device. Figure 1 The turnstile at the central passageway allows passage, and the user enters the restricted area.

[0077] If the channel face recognition server fails to recognize the face, the area face recognition server will perform face recognition. If the recognition is successful, the success message will be sent sequentially through the channel face recognition server and the front-end device. Figure 1 The turnstile at the central passageway allows passage, and the user enters the restricted area.

[0078] If the regional face recognition server fails to recognize the face, the central face recognition server will perform face recognition. If the recognition is successful, the success message will be sent sequentially through the regional face recognition server, the channel face recognition server, and the front-end device. Figure 1 The turnstile at the central passage allows users to enter the isolation area; if the recognition fails, the user is blocked and not allowed to pass.

[0079] In this embodiment, face recognition is a CPU-intensive service. The computing power should be: central face server > regional face server > channel face server. The central server has the best performance, but its recognition speed is relatively slow. When building a cluster, the service configuration of the central server should be given special consideration.

[0080] The specific network topology diagram for this embodiment can be found in [reference]. Figure 6 As shown, passenger information is collected on-site through content; passenger information collected via the app is sent to the central server via the external network. The central server, regional servers, channel servers, turnstiles, and other devices form a local area network that can communicate with each other.

[0081] This embodiment illustrates a user identification process performed by the system, specifically including the following steps:

[0082] First, users register using client-side applications such as mobile apps, during which their facial information is uploaded.

[0083] Secondly, after completing registration, users can enter the waiting area and choose to queue at a specific gate to pass through the next gate, such as... Figure 1 For example, if you choose to queue in area one, and the user appears within the range of the area camera corresponding to area one, the area camera will start collecting facial data and send it to the central face server for recognition. The recognition result will then be sent to the area face server.

[0084] Then, when a user arrives at a certain channel in the queue, the channel camera captures a single face image of the current user and sends the face information to the central face server for recognition. The central face server then sends the recognition result to the channel server.

[0085] Finally, once the identification is successful, the gate will allow passage.

[0086] In the description of this invention, it should be understood that the terms "coaxial," "bottom," "one end," "top," "middle," "other end," "upper," "side," "top," "inner," "outer," "front," "center," "both ends," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0087] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "setting," "connection," "fixing," "rotation," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal connection of two components or the interaction between two components. Unless otherwise explicitly limited, those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0088] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A dynamic face database based recognition system, characterized in that, The system comprises a channel camera, a region camera, a face tracking server, a channel face server, a region face server and a central face server, the face database of the central face recognition server is reduced according to different regions, the face database of each region is divided into a channel face recognition database according to channels, the face data is divided and reduced layer by layer, and a more accurate face database is formed, wherein: The area where the user is located is divided into a first detection area and a second detection area, the region camera is installed in the first detection area, and the channel camera is installed in the second detection area; The face tracking server is used for acquiring face information acquired by the channel camera and the region camera, and merging the face information acquired by the channel camera and the region camera; The central face server is used for storing face information left by a user during remote registration, and identifying face information collected by the face tracking server from the channel camera and the region camera, and sending the identification result to the channel face server and the region face server; The channel face server is used for storing the identification result of the channel camera sent by the central face server; The region face server is used for storing the identification result of the region camera sent by the central face server.

2. The system according to claim 1, wherein, The detection area comprises a plurality of second detection areas, each second detection area comprises a plurality of exit gate channels, and each gate is provided with a channel camera; one region camera is arranged in one second detection area.

3. The system according to claim 1, wherein, The central server only stores face information of valid users on the same day, and the valid users are users who have purchased tickets and whose flights have taken off on the same day.

4. The system according to claim 1, wherein, The region camera captures a group of face images with a face quality greater than a first threshold value, and the channel camera captures a face image with a face quality greater than a second threshold value, if the face image captured by the channel camera is in the group of face images captured by the region camera, it is determined that the similarity between the face image captured by the channel camera and the face image in the group of face images captured by the region camera is the largest, the information of the group of face images captured by the region camera is merged with the face information of the face image captured by the channel camera, and then the face information corresponding to the user is updated in the channel camera and the region camera.

5. The system according to claim 4, wherein, The second threshold value is greater than the first threshold value.

6. A dynamic face database-based recognition method, characterized in that, The face recognition is performed in the face recognition system based on the dynamic face database in claim 1, and specifically includes the following steps: The user registers remotely, collects face information of the user during registration, and stores the face information in the central face server; The channel camera performs face tracking on the user in the second detection area, and establishes a faceID for face information with a face image quality greater than a second threshold value and sends the faceID to the face tracking server; The region camera performs face tracking on the user in the first detection area, and establishes a faceID for face information with a face image quality greater than a first threshold value and sends the faceID to the face tracking server; The face tracking server includes face IDs corresponding to face information captured by the channel camera and the area camera, and determines whether the face IDs can be merged, and if so, merges and updates the face IDs corresponding to the face information of the channel camera and the area camera; The central face server compares the face information captured by the area camera with the face information registered by the user, and saves N images in the central face server that are closest to the face information captured by the area camera to the area face server; The central face server compares the face information captured by the channel camera with the face information registered by the user, and saves an image in the central face server that is closest to the face information captured by the area camera to the channel face server; The central face server determines whether the area face server already exists the face image according to the image saved by the channel face server, and if so, saves a face ID corresponding to a group of images with the largest similarity.

7. The method according to claim 6, wherein, The central server only stores face information of valid users on the day, and the valid users are users who have purchased tickets and whose flights have taken off on the day. 8.The method of claim 6, wherein, The second threshold is greater than the first threshold.

9. A computer device, comprising: The computer device includes: one or more processors; a memory for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement the identification method based on the dynamic face database as claimed in any one of claims 6-8.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The program is executed by the processor to implement the identification method based on the dynamic face database as claimed in any one of claims 6-8.

Images