Face feature library generation and use method for rail transit station based on face recognition

Abstract

This invention discloses a method for generating and using a facial feature database for rail transit stations based on facial recognition, comprising the following steps: (1) The central facial intelligent recognition management platform processes passenger facial information to generate personalized passenger feature information that can be used for downgraded services; (2) The station downgraded facial intelligent recognition management platform collects and processes the personalized passenger feature information that can be used for downgraded services to generate a station facial feature database; (3) The station facial feature database is used. This invention constructs a station facial feature information database that can be used for downgraded services. When the central facial intelligent recognition management platform's function declines or becomes ineffective, it can serve as a supplementary means to effectively improve the pass rate of active users at the station and the reliability of the facial recognition system.

Description

Technical Field

[0001] This invention relates to the field of intelligent urban rail transit technology, specifically to a method for generating and using a facial feature database for rail transit stations based on facial recognition. Background Technology

[0002] The "Outline for the Development of Smart Urban Rail Transit in China" released by the China Association of Urban Rail Transit also proposes the industry development goal of "by 2025, the widespread adoption of smart ticketing and inspection, real-name boarding, biometrics, contactless payment, and voice ticketing, seamless travel between cities, and the widespread application of the new model of smart ticketing and inspection integration".

[0003] With the increasing maturity of biometric technology, AI facial recognition technology has been widely and successfully applied in various industries such as public security, finance, and new retail. In urban rail transit, biometric features (such as faces) serve as an undeniable and highly secure personal biometric identity credential, completely freeing users from the constraints of tickets and mobile phones. This can also be considered a new mode of subway access, more advanced than QR code scanning technology, offering a better user experience. AI facial recognition technology has been gradually applied to urban rail transit automatic fare collection systems. A typical system architecture includes station terminal equipment, a central facial intelligent recognition management platform, and an internet ticketing management system. Station terminal equipment is primarily responsible for on-site facial image acquisition and passenger release management. The central facial intelligent recognition management platform is responsible for maintaining the facial database and comparing facial features, i.e., providing facial recognition services to the terminal equipment. The internet ticketing management system is responsible for passenger travel management.

[0004] The system architecture for AI facial recognition technology applications in the existing rail transit automatic fare collection industry is generally divided into a two-tier architecture. The first tier is the central facial recognition management platform, typically deployed in the network control center; the second tier is the station terminal equipment group, typically deployed in stations, providing facial recognition access services to passengers. Urban rail transit systems generally consist of multiple lines, with the network formed by ring networks of each line before converging to the central system. The reliability of the internal station network is higher than that of the network-level network. AI facial recognition technology primarily employs online real-time interactive technology, meaning that when station terminal equipment collects passenger facial information, it must promptly interact with the central facial recognition management platform to complete passenger facial recognition management. When the network experiences fluctuations or malfunctions, the reliability of the AI ​​facial recognition service will decrease or even fail. Therefore, it is necessary to add a station-level facial recognition management platform within the stations. Summary of the Invention

[0005] Purpose of the invention: In order to solve the problem that the reliability of AI face recognition services may decrease or even become unusable due to network oscillations or malfunctions in the two-level AI face recognition architecture, a method for generating and using a face feature database for rail transit stations based on face recognition is proposed.

[0006] Technical Solution: A station-level degraded facial recognition management platform is added to the second level of the AI ​​facial recognition technology application system architecture to manage station-level facial recognition services for active passengers. When the central facial recognition management platform becomes less functional or malfunctions, the station-level degraded facial recognition management platform takes over the station's facial recognition services, effectively improving the pass rate of active users and the reliability of the facial recognition system.

[0007] This invention provides a method for generating and using a facial feature database for rail transit stations based on facial recognition, comprising:

[0008] (1) The central facial recognition management platform processes passenger facial information to generate personalized passenger feature information that can be used for downgraded services.

[0009] The station terminal equipment collects passengers' facial information and sends it to the central facial intelligent recognition management platform for data processing. The uploaded information includes: device code, current time, built-in collection algorithm identifier, facial photo, etc.

[0010] The central facial recognition management platform incorporates multiple facial recognition computing clusters, each employing a different facial recognition algorithm. The facial image acquisition algorithm on the station terminal equipment must be matched with the facial recognition algorithm in the central facial recognition computing cluster.

[0011] The central facial recognition management platform extracts and compares features from collected facial photos and generates personalized passenger feature information that can be used for downgraded services. The relevant business rules are as follows:

[0012] 1) In the central facial recognition management platform, there are n facial database entries. Features are extracted from each passenger's registered photo, and basic feature information α is established. n Simultaneously, m active feature information α are generated for each passenger after self-learning. nm And set a similarity threshold δ.

[0013] 2) The central facial recognition management platform selects the corresponding algorithm cluster within the platform based on the algorithm identifier in the information uploaded by the device, and extracts features from the facial photo, resulting in γ. γ is then compared with the basic feature information α. n The similarity δ is obtained by comparison calculation. n Select the δ with the highest similarity. i And passenger characteristics α greater than δ iAs the preferred object, record the comparison result R1, which is composed of the passenger IDs corresponding to δ i , α i and α i .

[0014] 3) Map the passenger ID information corresponding to α i to obtain the active feature information α nm of this passenger. Use γ to compare with α nm again. Select k (k < m) records with a similarity greater than the similarity δ as the comparison results, and sort them from largest to smallest similarity, that is, R2, R3,... R k+1 .

[0015] 4) Combining the above, the passenger personalization feature information available for downgraded services consists of R1 and R2, R3,... R k+1 and can be represented by a matrix C:

[0016]

[0017] where i represents the passenger with the serial number i in the face database.

[0018] The central face intelligent recognition management platform feeds back the passenger personalization feature information available for downgraded services to the station terminal device. The passenger personalization feature information available for downgraded services includes: passenger ID, algorithm identifier, system time, passenger personalization feature information C i , etc.

[0019] (2) The station downgraded face intelligent recognition management platform performs aggregation processing on the passenger personalization feature information available for downgraded services to generate a station face feature library

[0020] When the line network is normal, the station terminal device collects the passenger's face information and sends it to the central face intelligent recognition management platform for data processing, and the central face intelligent recognition management platform feeds back the processing result to the station terminal device. The passenger personalization feature information available for downgraded services received by the station terminal device includes: whether it is a valid user, passenger ID, algorithm identifier, system time, passenger personalization feature information C i , etc.

[0021] When the station terminal device determines that the user is a valid user, first allow the passenger to pass, and then the device code, passenger ID, algorithm identifier, passenger personalization feature information C iThe information is uploaded to the station's downgraded facial recognition management platform, which then creates or updates the passenger's activity information. Simultaneously, the user information in the station's downgraded facial recognition management platform is collected and processed to generate a collection of information ∑C in the station's facial feature database. n .

[0022] (3) Use of the station's facial feature database

[0023] The station's downgraded facial recognition management platform incorporates a feature extraction and comparison algorithm that matches the algorithm used in the station's terminal equipment.

[0024] When the station terminal experiences a network malfunction with the central facial recognition management platform, the following services will be executed:

[0025] 1) The station terminal equipment collects passengers' facial information and sends it to the station's downgraded facial intelligent recognition management platform for data processing. The uploaded information includes: device code, current time, built-in collection algorithm identifier, facial photo, etc.

[0026] 2) The station's downgraded facial recognition management platform extracts and compares features from the collected facial photos;

[0027] The station's downgraded facial recognition management platform extracts features from passenger facial photos uploaded by station terminal devices, denoted as γ; and compares γ with C. i The basic characteristic information of active passengers is compared with that of α in R1. i Perform comparison and select the δ with the highest similarity. i And passenger characteristics α greater than δ i As a preferred release target.

[0028] To improve the accuracy of passenger facial recognition comparison, it is also necessary to further improve the accuracy of facial recognition comparison based on α. i The corresponding passenger ID information maps to the passenger's active characteristic information α. nm Using γ and α nm The comparison is repeated to calculate the number of records l with a similarity greater than similarity δ. When the number of records l is less than the set threshold, the facial photo information of the passenger collected this time needs to be temporarily saved and recorded. After the network is restored, the station's downgraded facial recognition management platform will send the collected passenger information to the central facial recognition management platform for secondary authentication and self-learning, and optimize its human database information.

[0029] Beneficial effects: Adding a station-level degraded facial recognition management platform to the second level of the AI ​​facial recognition technology application system architecture creates a station facial feature database for active users that can be used for degraded operations. When the central facial recognition management platform's functionality declines or malfunctions, it can serve as a supplementary measure, effectively improving the pass rate of active station users and the reliability of the facial recognition system. Attached Figure Description

[0030] Figure 1 Flowchart illustrating the generation and use of the station's facial feature database;

[0031] Figure 2 A method for generating personalized passenger information that can be used in downgraded services. Detailed Implementation

[0032] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0033] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of this application. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0034] It should be noted that the terminology used herein is for the purpose of describing particular implementations only and is not intended to limit the exemplary implementations according to this application.

[0035] Figure 1 The diagram shown illustrates the generation and use of a facial feature database for railway stations. This invention provides a method for generating and using a facial feature database for rail transit stations based on facial recognition, including:

[0036] (1) The central facial recognition management platform processes passenger facial information to generate personalized passenger feature information that can be used for downgraded business.

[0037] (1.1) The station terminal equipment collects passenger facial information and sends it to the central facial intelligent recognition management platform for data processing. The uploaded information includes: device code, current time, built-in collection algorithm identifier, and facial photo;

[0038] (1.2) The central face recognition management platform has multiple face recognition computing clusters built in. Each computing cluster uses a face recognition algorithm. The face photo acquisition algorithm on the station terminal equipment needs to be matched with the face recognition algorithm in the central face recognition computing cluster.

[0039] (1.3) The central facial recognition management platform extracts and compares features from the collected facial photos to generate personalized passenger feature information C for downgraded services. i ;

[0040] (1.4) The central facial recognition management platform will feed back the passenger's personalized feature information that can be used for downgraded business to the station terminal equipment.

[0041] (2) The station downgraded face intelligent recognition management platform collects and processes passenger personalized feature information that can be used for downgraded business and generates a station face feature database. Figure 2 The diagram shows a method for generating personalized passenger information that can be used for downgraded services.

[0042] (2.1) When the subway network is normal, the station terminal equipment collects passengers' facial information and sends it to the central facial recognition management platform for data processing. The central facial recognition management platform then feeds back the processing results to the station terminal equipment. The passenger personalization information received by the station terminal equipment that can be used for downgraded services includes: whether the user is valid, passenger ID, algorithm identifier, system time, and passenger personalization information C used for downgraded services. i wait.

[0043] (2.2) When the station terminal equipment determines that the user is a valid user, it first allows the passenger to pass, and then transmits the equipment code, passenger ID, algorithm identifier, and passenger personalized feature information C used for downgraded services. i The information is uploaded to the station's downgraded facial recognition management platform, which then creates or updates the passenger's activity information. Simultaneously, the user information in the station's downgraded facial recognition management platform is collected and processed to generate a collection of information ∑C in the station's facial feature database. n .

[0044] (3) Use of the station's facial feature database

[0045] (3.1) The station downgraded face intelligent recognition management platform has a built-in feature extraction and comparison algorithm that matches the acquisition algorithm in the station terminal equipment.

[0046] (3.2) When the network connection between the station terminal and the central facial recognition management platform is abnormal, the following services shall be executed:

[0047] (3.2.1) The station terminal equipment collects passengers’ facial information and sends it to the station’s downgraded facial intelligent recognition management platform for data processing. The uploaded information includes: device code, current time, built-in collection algorithm identifier, facial photo, etc.

[0048] (3.2.3) The station's downgraded facial recognition management platform extracts and compares features from the collected facial photos;

[0049] (3.2.3) The station-level downgraded facial recognition management platform extracts features from passenger facial photos uploaded by station terminal devices, denoted as γ; and compares γ with C. i The basic characteristic information of active passengers is compared with that of α in R1. i Perform comparison and select the δ with the highest similarity. i And passenger characteristics α greater than δ i As a preferred release target.

[0050] (3.3) To improve the accuracy of passenger facial recognition comparison, it is also necessary to further improve the accuracy of facial recognition comparison based on α. i The corresponding passenger ID information maps to the passenger's active characteristic information α. nm Using γ and α nm The comparison is repeated to calculate the number of records l with a similarity greater than similarity δ. When the number of records l is less than the set threshold, the facial photo information of the passenger collected this time needs to be temporarily saved and recorded. After the network is restored, the station's downgraded facial recognition management platform will send the collected passenger information to the central facial recognition management platform for secondary authentication and self-learning, and optimize its human database information.

[0051] In this embodiment of the invention, step (1.3) specifically includes:

[0052] (S1) In the central facial recognition management platform, there are n facial database entries. For each passenger's registered photo, features are extracted and basic feature information α is established. n Simultaneously, m active feature information α are generated for each passenger after self-learning. nm And set a similarity threshold δ.

[0053] (S2) The central facial recognition management platform selects the corresponding algorithm cluster within the platform based on the algorithm identifier in the information uploaded by the device, and extracts features from the facial photo, resulting in γ. γ is then compared with the basic feature information α. n The similarity δ is obtained by comparison calculation. n Select the δ with the highest similarity. i And passenger characteristics α greater than δ i As the preferred candidate, the comparison result R1 is recorded, where R1 is determined by δ. i α i and α i The corresponding passenger ID composition.

[0054] (S3) According to α i The corresponding passenger ID information maps to the passenger's active characteristic information α. nm Using γ and α nmCompare again, select k (k < m) records with similarity greater than similarity δ as the comparison result, and sort them in descending order of similarity, i.e., R2, R3, ... R k+1 .

[0055] (S4) Combining the above, the passenger personalization feature information available for the downgraded service consists of R1 and R2, R3, ... R k+1 and can be represented by matrix C:

[0056]

[0057] where i represents the passenger with serial number i in the face database.

Claims

1. A method for generating and using a facial feature database for rail transit stations based on facial recognition, characterized in that, Includes the following steps: (1) The central facial recognition management platform processes passenger facial information to generate personalized passenger feature information that can be used for downgraded services; (2) The station downgraded face intelligent recognition management platform collects and processes passenger personalized feature information that can be used for downgraded business, and generates a station face feature database; (3) Use the station's facial feature database; Step (1) specifically includes: (1.1) The station terminal equipment collects passenger facial information and sends it to the central facial intelligent recognition management platform for data processing. The uploaded information includes: device code, current time, built-in collection algorithm identifier, and facial photo; (1.2) The central face intelligent recognition management platform has multiple face recognition computing clusters built in. Each computing cluster uses a face recognition algorithm. The face photo acquisition algorithm on the station terminal equipment needs to be matched with the face recognition algorithm in the central face recognition computing cluster. (1.3) The central facial recognition management platform extracts and compares features from the collected facial photos to generate personalized passenger feature information for downgraded services. ; (1.4) The central facial recognition management platform feeds back the personalized passenger feature information that can be used for downgraded services to the station terminal equipment; step (1.3) specifically includes: (1.3.1) A facial database is set up in the central facial intelligent recognition management platform. For each passenger's registered photo, features are extracted and basic feature information is established. At the same time, it also establishes a self-learning system for each passenger. Active feature information And set a similarity threshold. ; (1.3.2) The central facial recognition management platform selects the corresponding algorithm cluster within the platform based on the algorithm identifier in the information uploaded by the device, and extracts features from the facial photo. The result is... ,Will Basic feature information Perform comparison calculations to obtain similarity. Select the one with the highest similarity and greater than Passenger characteristics As the preferred candidate, the comparison results were recorded. , Depend on , and The corresponding passenger ID composition; (1.3.3) According to the corresponding passenger ID information, map out the active feature information of the passenger , and use to compare again, and select the records with a similarity greater than the similarity as the comparison result, where k < m, and sort them in descending order of similarity, that is ; (1.3.4) In conjunction with the above, the passenger personalization information that can be used for downgrade services is provided by and The composition can be represented by matrix C: Where i represents the passenger with serial number i in the face database.

2. The method for generating and using a facial feature database for rail transit stations based on facial recognition according to claim 1, characterized in that, The passenger personalized feature information that can be used for downgrade services includes: passenger ID, algorithm identifier, system time, and passenger personalized feature information used for downgrade services. .

3. The face feature library generation and use method based on face recognition in a rail transit station according to claim 1, characterized in that, Step (2) specifically includes: (2.1) When the rail network is normal, the station terminal equipment collects passengers' facial information and sends it to the central facial intelligent recognition management platform for data processing. The central facial intelligent recognition management platform then feeds back the processing results to the station terminal equipment. The passenger personalized feature information received by the station terminal equipment that can be used for downgraded services includes: whether the user is valid, passenger ID, algorithm identifier, system time, and passenger personalized feature information used for downgraded services. ; (2.2) When the station terminal equipment determines that the user is a valid user, it first allows the passenger to pass, and then transmits the equipment code, passenger ID, algorithm identifier, and passenger personalized feature information used for downgraded services. The information is uploaded to the station's downgraded facial recognition management platform, which then creates or updates the passenger's activity information. Simultaneously, the platform aggregates and processes user information from the station's downgraded facial recognition management platform to generate a collection of information for the station's facial feature database. .

4. The face feature library generation and use method based on face recognition in a rail transit station according to claim 1, characterized in that, Step (3) specifically includes: (3.1) The station-level downgraded facial recognition management platform has a built-in feature extraction and comparison algorithm that matches the acquisition algorithm in the station terminal equipment; (3.2) When the network between the station terminal and the central facial recognition management platform is abnormal, the station terminal equipment collects the passengers' facial information and sends it to the station's downgraded facial recognition management platform for data processing; (3.3) In order to improve the accuracy of passenger facial recognition comparison, it is also necessary to... The corresponding passenger ID information maps to the passenger's activity characteristics. ,use A second comparison was performed, and the calculated similarity score was greater than the previous one. Number of records When the number of records When the value is less than the set threshold, the facial image information of the passenger collected this time needs to be temporarily recorded. After the network is restored, the station's downgraded facial recognition management platform will send the collected passenger information to the central facial recognition management platform for secondary authentication and self-learning, and optimize its human resource database information.

5. The face feature library generation and use method based on face recognition in a rail transit station according to claim 4, characterized in that, Step (3.2) specifically includes: (3.2.1) The station terminal equipment collects passengers’ facial information and sends it to the station’s downgraded facial intelligent recognition management platform for data processing. The uploaded information includes: device code, current time, built-in collection algorithm identifier, and facial photo; (3.2.3) The station's downgraded facial recognition management platform extracts and compares features from the collected facial photos; (3.2.3) The station-level downgraded facial recognition management platform extracts features from passenger facial photos uploaded by station terminal devices, described as follows: ;Will and Compare the basic characteristic information of active passengers with, that is, with In Compare and select the one with the highest similarity. And greater than Passenger characteristics As a preferred release target.

Images