School friend registration authentication method and system based on certificate photo recognition

Abstract

The invention discloses a school friend registration authentication method and system based on identification photo recognition, and the method comprises the steps: firstly obtaining an identification photo of a registrant and a real-time selfie photo containing the face of the registrant, carrying out the preprocessing of the obtained identification photo and the real-time selfie photo, and collecting a corresponding face image only containing a face; secondly, face feature extraction with discrimination is carried out on the collected face images, and similarity calculation analysis is carried out on the face image in the certificate photo and the face image in the real-time self-shot photo according to the extracted face features; thirdly, keyword extraction is carried out on the obtained certificate photo, and the extracted keywords are compared with registration information input by a registrant; and finally, determining a registration authentication result according to a face image similarity calculation result and a registration information comparison result. According to the scheme, the authenticity of the registration information can be efficiently confirmed, and the overall registration speed and accuracy can be effectively improved under the condition that the authenticity of the registration information is ensured.

Description

Technical Field

[0001] The present invention relates to identity registration and authentication technology, and more particularly to alumni registration and authentication technology for identifying alumni identities based on ID photos. Background Art

[0002] Traditional alumni registration and authentication methods mainly rely on users to manually fill in personal information, including but not limited to key data such as name, contact information, ID number, graduation year, etc. In the specific implementation of this manual method, the software functional modules involved mainly include a user interface, a data input module, a data storage module, and a manual review module. Among them, after the user fills in the information through the interface, the data input module is responsible for collecting this information and storing it in the database. Subsequently, the manual review module will check the submitted information one by one to ensure the authenticity of the information.

[0003] However, there are many problems with the traditional method. First of all, manually filling in information is not only time-consuming and laborious, but also prone to information errors or omissions due to human negligence. Secondly, the manual review process is cumbersome and inefficient. Especially when faced with a large number of registration requests, the workload of the reviewers increases significantly. More critically, the traditional method is difficult to effectively prevent impersonation registration or false registration, posing a great challenge to the management of the alumni association and the authenticity of alumni information.

[0004] In response to this, in recent years, a registration method based on face recognition technology has emerged to improve the efficiency and accuracy of registration and authentication. This method is mainly achieved through the cooperation of a face recognition functional module, an information comparison functional module, and a comparison result output functional module. Among them, the face recognition functional module generally uses advanced image recognition algorithms to extract the facial features of alumni and compare them with the pre-established personnel information photo database. The information comparison functional module is specifically responsible for judging the matching degree between the extracted features and the information in the database, and finally the result output module gives the authentication result.

[0005] Although the face recognition registration method has improved the automation degree and accuracy of authentication to a certain extent, there are still the following problems in the actual application process:

[0006] (1) Requirement for a personnel information photo database: In practical applications, it is very difficult to build a comprehensive and accurate personnel information photo database. On the one hand, the collection, collation, and storage of historical data require a large amount of time and resources. On the other hand, there is also a possibility of omission or loss of historical data.

[0007] (2)Low accuracy of face recognition based on student ID card information: The photos on the student ID card are often old and affected by factors such as shooting conditions and resolution. There may be significant differences between the facial features on the card and the current actual appearance. This makes it difficult for face recognition algorithms to achieve high accuracy during comparison, thus affecting the accuracy of authentication.

[0008] In summary, although the original face recognition registration method has improved the efficiency and accuracy of alumni registration and authentication to a certain extent, there are still many technical and implementation problems.

[0009] Therefore, providing a more efficient, accurate and convenient alumni registration and authentication solution is an urgent problem to be solved in this field. Summary of the Invention

[0010] Aiming at the deficiencies in the authentication efficiency and accuracy of the existing alumni registration and authentication solutions, the purpose of the present invention is to provide an alumni registration and authentication solution based on ID photo recognition. This authentication solution realizes the identity authentication of the registrant based on the ID photo of the registrant and the real-time self-taken photo of the registrant, and can effectively improve the overall registration speed and accuracy while ensuring the authenticity of the registration information.

[0011] To achieve the above purpose, the present invention provides an alumni registration and authentication method based on ID photo recognition, including:

[0012] (1) Obtain the ID photo of the registrant and the real-time self-taken photo containing the face of the registrant, and preprocess the obtained ID photo and real-time self-taken photo respectively to collect the corresponding face images containing only the face.

[0013] (2) Extract distinguishable face features from the collected face images respectively, and calculate and analyze the similarity between the face image in the ID photo and the face image in the real-time self-taken photo according to the extracted face features.

[0014] (3) Extract key text from the obtained ID photo, and compare the extracted key text with the registration information input by the registrant.

[0015] (4) Determine the registration and authentication result according to the face image similarity calculation result in step (2) and / or the registration information comparison result in step (3).

[0016] In some embodiments of the present invention, when preprocessing the ID photo in step (1), first, the position and size information of the face image are extracted from the ID photo; then, based on the extracted position and size information of the face image, the face image in the ID photo is recognized. Taking the facial feature points of the frontal face image in the ID photo as reference points, the corresponding contour points in the ID photo are detected, and then the face image in the ID photo is mapped to the reference point position by using affine transformation, thereby obtaining a face image containing only the face.

[0017] In some embodiments of the present invention, when preprocessing the real-time selfie photo in step (1), first, the real-time selfie photo is detected for reshooting, and then the position and size information of the face image are extracted from the real-time selfie photo; then, based on the extracted position and size information of the face image, the face image in the real-time selfie photo is recognized. Taking the facial feature points of the frontal face image in the real-time selfie photo as reference points, the corresponding contour points in the real-time selfie photo are detected, and then the face image in the real-time selfie photo is mapped to the reference point position by using affine transformation, thereby obtaining a face image containing only the face.

[0018] In some embodiments of the present invention, in step (2), discriminative face feature extraction is performed on the face image based on the dual-branch LeNet-5 convolutional neural network.

[0019] In some embodiments of the present invention, in step (2), the Contrastive Loss function is used to optimize the fusion network.

[0020] To achieve the above object, the present invention provides an alumni registration and authentication system based on ID photo recognition. The alumni registration and authentication system includes an image acquisition and preprocessing unit, a feature extraction unit, an analysis and comparison unit, and a comprehensive determination unit.

[0021] The image acquisition and preprocessing unit is configured to be able to obtain the ID photo of the registrant and the real-time selfie photo containing the face of the registrant, and preprocess the obtained ID photo and real-time selfie photo respectively to collect the corresponding face images containing only the face.

[0022] The feature extraction unit is configured to perform data interaction with the image acquisition and preprocessing unit, and be able to perform discriminative face feature extraction on the collected face images respectively, and be able to extract key texts from the obtained ID photos.

[0023] The analysis and comparison unit is configured to interact with the feature extraction unit, and can calculate and analyze the similarity between the face image in the ID photo and the face image in the real-time selfie photo according to the face features extracted by the feature extraction unit; and compare the key text extracted by the feature extraction unit with the registration information input by the registrant.

[0024] The comprehensive determination unit is configured to interact with the analysis and comparison unit, and can determine the registration authentication result according to the face image similarity calculation result and / or the registration information comparison result in the analysis and comparison unit.

[0025] In some embodiments of the present invention, the image acquisition and preprocessing unit includes:

[0026] An image acquisition module, which is configured to acquire the ID photo of the registrant and the real-time selfie photo including the face of the registrant.

[0027] A face image detection module, which is configured to interact with the image acquisition module, and can extract the position and size information of the corresponding face image from the acquired ID photo or real-time selfie photo.

[0028] A face image extraction module, which is configured to interact with the face image detection module, and can identify the face image in the ID photo or real-time selfie photo based on the face image position and size information extracted by the face image detection module. Taking the facial feature points of the front face image in the ID photo or real-time selfie photo as the reference points, the corresponding contour points in the ID photo or real-time selfie photo are detected, and then the face image in the ID photo or real-time selfie photo is mapped to the reference point position by affine transformation, thereby obtaining a corresponding face image containing only the face.

[0029] In some embodiments of the present invention, the image acquisition and preprocessing unit further includes a live detection module, which is configured to interact with the image acquisition module and can perform a reshooting detection on the acquired real-time selfie photo.

[0030] In some embodiments of the present invention, a fusion network model is constructed in the feature extraction unit, and a dual-branch LeNet-5 convolutional neural network is used to perform discriminative face feature extraction on the face image respectively.

[0031] In some embodiments of the present invention, the Contrastive Loss function is also used to optimize the fusion network in the feature extraction unit.

[0032] The alumni registration and authentication solution based on ID photo recognition provided by the present invention innovatively realizes the identity authentication of the registrant based on the ID photo of the registrant and the real-time self-taken photo of the registrant, can efficiently confirm the authenticity of the registration information, and can effectively improve the overall registration speed and accuracy while ensuring the authenticity of the registration information.

[0033] The alumni registration and authentication solution based on ID photo recognition provided by the present invention has the following excellent effects compared with the prior art:

[0034] (1) It can ensure the authenticity of the identity of the registrant: The solution of the present invention collects the ID photo and real-time self-taken photo of the user, uses the live detection technology to exclude the interference of non-living bodies, and ensures that the registered individual is real; at the same time, it also accurately compares the facial features of the user with the features in the ID photo to verify the authenticity of the user's identity.

[0035] (2) Improve the efficiency of alumni registration: Compared with the traditional manual filling of information registration method, which is cumbersome and time-consuming, and is also prone to information errors or omissions due to human negligence, the solution of the present invention greatly shortens the time required for the registration process by automatically processing the ID photo and real-time self-taken photo of the registrant, and at the same time reduces the possibility of human errors; users only need to simply upload the required materials, and the subsequent identification and verification work can be automatically completed, thus significantly improving the registration efficiency.

[0036] (3) Improve the accuracy and adaptability of face recognition technology: The solution of the present invention significantly improves the accuracy and adaptability of face recognition by introducing ID photo comparison, supplemented by live detection technology and feature extraction and matching methods; in addition, the solution of the present invention also combines auxiliary means such as name comparison to further enhance the reliability of registration and authentication. Description of the Drawings

[0037] The following further illustrates the present invention in conjunction with the drawings and specific embodiments.

[0038] Figure 1 It is the flow chart of alumni registration and authentication based on ID photo recognition in the present invention;

[0039] Figure 2 It is the composition example diagram of the alumni registration and authentication system based on ID photo recognition in the present invention;

[0040] Figure 3 It is the example diagram of live detection for images in the example of the present invention;

[0041] Figure 4 It is the example diagram of noise reduction processing for images in the example of the present invention. Detailed Embodiments

[0042] In order to make the technical means, creative features, achieved purposes and effects of the present invention easy to understand, the present invention will be further elaborated below in conjunction with specific illustrations.

[0043] During the alumni registration and authentication process, it is necessary to ensure the authenticity and accuracy of the registration and authentication information. Therefore, during the registration process, it is necessary to determine that the registrant is himself / herself. In this regard, the present invention abandons the existing registration and authentication scheme based on face recognition using student ID card photos, and innovatively realizes the real identity authentication of the registrant through the ID card photo of the registrant and the real-time self-taken photo of the registrant, thereby efficiently and accurately completing the alumni registration and authentication process.

[0044] Accordingly, the present invention provides an alumni registration and authentication method based on ID card photo recognition, as Figure 1 shown, this alumni registration and authentication method is mainly realized by the sequential cooperation of four stages: image acquisition and preprocessing, feature extraction, feature analysis and comparison, and comprehensive judgment.

[0045] (1) Image acquisition and preprocessing stage,

[0046] In this stage, the ID card photo of the registrant and the real-time self-taken photo containing the face of the registrant are obtained, and the obtained ID card photo and real-time self-taken photo are respectively preprocessed to collect the corresponding face images containing only the face.

[0047] (2) Feature extraction stage,

[0048] In this stage, discriminative face features are extracted respectively for the face images collected in the image acquisition and preprocessing stage;

[0049] In this stage, key text is also extracted from the ID card photo collected in the image acquisition and preprocessing stage.

[0050] (3) Feature analysis and comparison stage,

[0051] In this stage, according to the face features extracted in the feature extraction stage, feature matching is performed on the face image in the ID card photo and the face image in the real-time self-taken photo, and based on this, similarity calculation and analysis are carried out to form a similarity score value that can reflect the similarity between the face image in the ID card photo and the face image in the real-time self-taken photo.

[0052] In this stage, the key text extracted in the feature extraction stage is also compared and analyzed with the registration information input by the registrant to form a comparison result.

[0053] (4) Comprehensive judgment stage,

[0054] In this stage, the registration and authentication result is finally determined according to the face image similarity calculation result and the registration information comparison result in the feature analysis and comparison stage.

[0055] The following describes in detail the specific implementation scheme and corresponding technical features of each stage when implementing this alumni registration and authentication method based on ID photo recognition.

[0056] The alumni registration and authentication method obtains the ID photo of the registrant during the image acquisition and preprocessing stage, which can be formed by directly photographing the registrant's ID through a camera component, or directly obtaining an existing ID photo uploaded by the registrant.

[0057] In order to ensure the authenticity of the photo, it is preferred in this solution to directly photograph the registrant's ID card using a camera assembly to form the registrant's ID card photo.

[0058] The real-time selfie photo containing the registrant's face is obtained by directly photographing the registrant's head portrait through a camera component (such as the mobile phone used during registration).

[0059] In the image acquisition and preprocessing stage, this alumni registration and authentication method preprocesses the obtained certificate photo through a face image detection step and a face image extraction step.

[0060] Specifically, firstly, face image detection is performed on the acquired ID photo, that is, the position and size information of the face image is extracted from the ID photo to cooperate with the subsequent face alignment operation and feature extraction operation.

[0061] Next, the facial image in the ID photo is identified and extracted based on the extracted facial image position and size information. The facial feature points of the frontal facial image in the ID photo are used as reference points to detect the corresponding contour points in the ID photo. Then, affine transformation is used to correspond the facial image in the ID photo to the reference point position, thereby obtaining a facial image containing only the face.

[0062] The facial image containing only the human face in the ID photo thus extracted can be adapted to the facial feature extraction in the subsequent feature extraction stage to ensure the accuracy of the feature extraction.

[0063] In the image acquisition and preprocessing stage, when preprocessing the acquired real-time selfie photos, the alumni registration and authentication method is implemented by coordinating the photo copy detection step, the face image detection step and the face image extraction step in sequence.

[0064] Specifically, first, a copy detection is performed on the acquired real-time selfie photo. This step analyzes the acquired real-time selfie photo through liveness detection technology to determine whether the selfie is based on a real face, rather than a copy of other portrait photos or videos.

[0065] Meanwhile, this solution directly associates the detection result of this reshooting detection step with the comprehensive judgment stage, that is, if the detection result is reshooting, it directly enters the comprehensive judgment stage to end the entire registration and authentication process.

[0066] Through this step, real human faces can be effectively distinguished from non-living human faces such as photos, thus avoiding the risk of being fraudulently attacked.

[0067] Next, the position and size information of the human face image are extracted from the real-time self-taken photo that has passed the reshooting detection to cooperate with the subsequent face alignment operation and feature extraction operation.

[0068] Finally, based on the extracted position and size information of the human face image, the human face image in the real-time self-taken photo is recognized. Taking the facial feature points of the frontal human face image in the real-time self-taken photo as reference points, the corresponding contour points in the real-time self-taken photo are detected, and then the human face image in the real-time self-taken photo is mapped to the reference point position by using affine transformation, thereby obtaining a human face image containing only the human face.

[0069] As a further illustration, in the reshooting detection step of the solution of the present invention, Haar wavelet basis is preferably used to perform wavelet decomposition on the real-time self-taken photo, thereby realizing efficient reshooting detection and enabling this solution to be applicable to real-time scenarios such as alumni registration.

[0070] The human face image containing only the human face extracted from the ID photo can be adapted to the human face feature extraction in the subsequent feature extraction stage to ensure the accuracy of feature extraction.

[0071] In the feature extraction stage of this alumni registration and authentication method, a fusion network model is preferably constructed, and then discriminative human face features are extracted from the human face image based on the dual-branch LeNet-5 convolutional neural network.

[0072] As a further illustration, the Contrastive Loss function is also used in this solution to optimize the fusion network, thereby improving the recognition efficiency.

[0073] Specifically, in the solution of the present invention, the Contrastive Loss function is used to train the features of the human face recognition area of the portrait. By configuring appropriate positive and negative sample pairs and adjusting the parameter m value in the Contrastive Loss function, the features of the human face area of the ID photo portrait and the corresponding human face area features of the self-taken photo of the same person are made more similar, thereby improving the recognition efficiency in the subsequent portrait comparison.

[0074] As a further illustration, when extracting key text from the ID photo, the extracted key text information includes key information such as name and ID number, but is not limited thereto.

[0075] As a further illustration, in the feature analysis and comparison stage of this solution, when calculating and analyzing the similarity between the face image in the ID photo and the face image in the real-time selfie photo, specifically, the face features of the face image in the ID photo extracted are matched and compared with the face features of the face image in the real-time selfie photo. Based on this, the similarity calculation and analysis are carried out to form the corresponding similarity score value, and this similarity score value is compared with a preset threshold to determine whether the face image in the real-time selfie photo and the face image in the ID photo are of the same person. As an example, if the similarity score value is higher than the threshold, it is determined that the face image in the real-time selfie photo and the face image in the ID photo are of the same person; if it is lower than the threshold, it is determined that the face image in the real-time selfie photo and the face image in the ID photo are not of the same person.

[0076] As a further illustration, in the feature analysis and comparison stage of this solution, when comparing and analyzing the key text extracted with the registration information input by the registrant, the real name and / or ID number injected by the registrant are obtained and compared with the name and / or ID number extracted. Only when the two are consistent, a positive comparison result is formed, and thus the final comparison result is formed.

[0077] In the comprehensive judgment stage of this alumni registration and authentication method, for the determination mode of the registration and authentication result, it can be determined according to actual needs and is not limited here.

[0078] As an example, in the comprehensive judgment stage, the face image similarity calculation result and the registration information comparison result in the feature analysis and comparison stage are obtained simultaneously for comprehensive determination: if both pass the verification, the registration and authentication are passed; otherwise, the registration and authentication are rejected, and the corresponding error information is given.

[0079] It should be noted here that the specific registration and authentication determination mode is not limited to this.

[0080] As can be seen from the above, the alumni registration and authentication method based on ID photo recognition given by the present invention not only improves the accuracy of authentication, but also enhances the security, effectively avoiding the risk of identity theft for registration in the existing technology, and at the same time preventing fraudulent attacks. It can enable registered users to complete alumni registration and authentication efficiently and accurately, providing a solid guarantee for the management of the alumni association and the authenticity of alumni information.

[0081] In addition, the comprehensive determination mechanism introduced in the alumni registration and authentication method based on ID photo recognition given by the present invention, which combines two means of face determination and name comparison for determination, further improves the accuracy and reliability of authentication, and reduces the misjudgment and missed judgment caused by a single verification factor.

[0082] For the alumni registration and authentication method based on ID photo recognition given in this example solution, in specific applications, it can form a corresponding software program to form a corresponding alumni registration and authentication system based on ID photo recognition. When this software program runs, it will execute the above-mentioned alumni registration and authentication method based on ID photo recognition, and at the same time be stored in a corresponding storage medium for the processor to retrieve and execute.

[0083] See Figure 2 , and the resulting alumni registration and authentication system 100 based on ID photo recognition mainly includes four functional units in terms of function: an image acquisition and preprocessing unit 110, a feature extraction unit 120, a feature similarity analysis and comparison unit 130, and a comprehensive determination unit 140.

[0084] The image acquisition and preprocessing unit 110 in this system is specifically configured to be able to acquire the ID photo of the registrant and the real-time self-taken photo containing the face of the registrant, and preprocess the acquired ID photo and real-time self-taken photo respectively to collect the corresponding face images containing only the face.

[0085] The feature extraction unit 120 in this system is configured to perform data interaction with the image acquisition and preprocessing unit 110, and be able to perform discriminative face feature extraction on the collected face images respectively, and be able to perform key text extraction on the acquired ID photo.

[0086] The feature analysis and comparison unit 130 in this system is configured to perform data interaction with the feature extraction unit 120, and be able to perform similarity calculation and analysis on the face images in the ID photo and the face images in the real-time self-taken photo according to the face features extracted by the feature extraction unit 120; and compare the key text extracted by the feature extraction unit with the registration information input by the registrant;

[0087] The comprehensive determination unit 140 in this system is configured to perform data interaction with the feature analysis and comparison unit 130, and be able to determine the registration and authentication result according to the face image similarity calculation result and / or the registration information comparison result in the feature analysis and comparison unit 130.

[0088] Here, for the corresponding functional units in this system, a specific composition scheme is further given.

[0089] Combined with Figure 2 As shown, the image acquisition and preprocessing unit 110 in this system is specifically composed of an image acquisition module 111, a live detection module 112, a face image detection module 113, and a face image extraction module 114 cooperating with each other.

[0090] Specifically, the image acquisition module 111 therein is configured to collect and acquire the ID photo of the registrant and the real-time self-taken photo containing the face of the registrant.

[0091] Here, no specific implementation scheme for acquisition is defined, and it can be determined according to actual requirements.

[0092] The live detection module 112 in this unit is configured to interact with the image acquisition module 111, and is specifically used for detecting the reshooting of the real-time self-taken photo of the registrant's face acquired by the image acquisition module 111.

[0093] As a further illustration, the live detection module 112 analyzes the acquired real-time self-taken photo through live detection technology, thereby determining whether the self-taken photo is based on a real human face, rather than reshooting other portrait photos or videos.

[0094] As an example, the live detection module 112 preferably uses Haar wavelet basis to perform wavelet decomposition on the real-time self-taken photo, thereby realizing reshooting detection.

[0095] As a further illustration, the live detection module 112 is also configured to interact with the comprehensive determination unit 140, and can directly associate the detection result with the comprehensive determination unit 140, that is, if the detection result is reshooting, it will directly jump to the comprehensive determination unit 140 to end the entire registration and authentication process.

[0096] The face image detection module 113 in this unit is configured to interact with the image acquisition module 111 and the live detection module 112, and can extract the position and size information of the corresponding face image from the acquired ID photo or the real-time self-taken photo after live detection, so as to cooperate with subsequent face alignment operations and feature extraction operations.

[0097] The face image extraction module 114 in this unit is configured to interact with the face image detection module 113, and can identify the face image in the ID photo or the real-time self-taken photo based on the face image position and size information extracted by the face image detection module 113. Taking the facial feature points of the frontal face image in the ID photo or the real-time self-taken photo as the reference points, the corresponding contour points in the ID photo or the real-time self-taken photo are detected, and then the face image in the ID photo or the real-time self-taken photo is mapped to the reference point position by using affine transformation, thereby obtaining a corresponding face image containing only the face.

[0098] The feature extraction unit 120 in this system is specifically composed of a cooperation between the face feature extraction module 121 and the key text extraction module 122.

[0099] Among them, the face feature extraction module 121 is specifically configured to interact with the image acquisition and preprocessing unit 110, and can perform discriminative face feature extraction on the face images acquired by the image acquisition and preprocessing unit 110 respectively.

[0100] Furthermore, the face feature extraction module 121 extracts discriminative face features from face images by constructing a fusion network model and using a dual-branch LeNet-5 convolutional neural network.

[0101] As a further illustration, the face feature extraction module 121 also optimizes the fusion network using the Contrastive Loss function.

[0102] As an example, the face feature extraction module 121 introduces the Contrastive Loss function to train the features of the face recognition area of the portrait. By configuring appropriate positive and negative sample pairs and adjusting the parameter m value in the Contrastive Loss function, the face area features of the portrait in the ID photo and the corresponding face area features of the self-taken photo of the same person are made more similar, thus improving the recognition efficiency in subsequent portrait comparisons.

[0103] The key text extraction module 122 is specifically configured to interact with the image acquisition and preprocessing unit 110, be able to obtain the ID photos collected by the image acquisition and preprocessing unit 110, and extract key text from the ID photos.

[0104] As a further illustration, when the key text extraction module 122 extracts key text from the ID photo, the extracted key text information preferably includes key information such as name and ID number, but is not limited thereto.

[0105] The feature analysis and comparison unit 130 in this system is specifically composed of a similarity calculation and analysis module 131 and a registration information comparison module 132 working together.

[0106] Among them, the similarity calculation and analysis module 131 is configured to interact with the feature extraction unit 120, be able to perform feature matching on the face image in the ID photo and the face image in the real-time self-taken photo according to the face features extracted by the feature extraction unit 120, and perform similarity calculation and analysis accordingly, forming a similarity score value that can reflect the similarity between the face image in the ID photo and the face image in the real-time self-taken photo.

[0107] As a further illustration, the similarity calculation and analysis module 131 specifically matches and compares the face features of the face image in the extracted ID photo with the face image features in the extracted real-time self-taken photo, performs similarity calculation and analysis accordingly to form a corresponding similarity score value, and compares the similarity score value with a preset threshold to determine whether the face image in the real-time self-taken photo and the face image in the ID photo are the same person.

[0108] The registration information comparison module 132 in this unit is configured to interact with the feature extraction unit 120, and can compare and analyze the key text extracted by the feature extraction unit 120 with the registration information input by the registrant to form a comparison result.

[0109] As a further illustration, when the registration information comparison module 132 compares and analyzes the extracted key text with the registration information input by the registrant, it obtains the real name and / or ID number injected by the registrant, and compares it with the extracted name and / or ID number. Only when the two are consistent, a positive comparison result is formed, thereby forming the final comparison result.

[0110] For the alumni registration and authentication scheme based on ID photo recognition given in the present invention, the following specific examples are used to further illustrate its implementation process and corresponding technical features.

[0111] When the alumni registration and authentication scheme based on ID photo recognition is applied and implemented, first, a corresponding alumni registration and authentication system is constructed based on the foregoing scheme. The alumni registration and authentication system can be integrated with existing alumni management software or systems to complete the corresponding alumni registration and authentication process.

[0112] At the same time, the alumni registration and authentication system can be presented in the form of an APP, PC client software or WEB plugin, but is not limited thereto. The alumni registration and authentication system can run on any intelligent electronic device.

[0113] Here, taking the operation of the alumni registration and authentication system APP (hereinafter referred to as the system) in an intelligent terminal as an example, the implementation process of completing alumni registration is described. Combining Figure 1 As shown, the entire registration and authentication process is as follows:

[0114] (1) Image acquisition:

[0115] The registered user first takes a photo of his / her ID with the camera of the intelligent terminal to form an ID photo and uploads the ID photo. According to needs, he / she can also directly obtain an existing ID photo through the intelligent terminal for uploading; at the same time, take a real-time selfie photo containing his / her face with the camera of the intelligent terminal and upload it to the system as well, and it is collected by the image acquisition module in the image acquisition and preprocessing unit.

[0116] (2) Replay detection:

[0117] The system first analyzes the uploaded real-time selfie photo by the live detection module in the image acquisition and preprocessing unit to determine whether it is a selfie based on a real face rather than a replay of someone else's portrait photo or video: if the detection result is a replay, an error message is directly returned and the registration process is terminated.

[0118] The live detection module in this system uses Haar wavelet basis to perform wavelet decomposition on self-taken photos for detecting reshot photos.

[0119] As Figure 3 shown, where Figure 3 (a) is a real self-taken photo, Figure 3 (b) is the image after brightness processing of the real self-taken photo, Figure 3 (c) is a reshot picture, Figure 3 (d) is the image after brightness processing of the reshot picture.

[0120] It can be seen by comparison that there are obvious differences between the real face self-taken photo and the reshot face photo in the high-frequency sub-band region after the second-level decomposition of Haar wavelet.

[0121] In addition, the mother wavelet of Haar wavelet here is:

[0122]

[0123] (2) Face detection:

[0124] The face image detection module in the system extracts the position and size of the face image in the ID photo and the self-taken photo respectively to prepare for subsequent face alignment and feature extraction.

[0125] The face image detection module in this system specifically uses the weighted average method to convert the ID portrait photo and the self-taken photo into grayscale images, thereby maintaining the main features of the images and removing noise interference.

[0126] Specifically, since the human eye is most sensitive to green and least sensitive to blue, the grayscale image is obtained by weighted averaging the RGB three components according to the following formula:

[0127] Gray(i,j) = 0.299 * R(i,j) + 0.578 * G(i,j) + 0.114B * (i,j);

[0128] On this basis, the Laplace edge detection operator is further used for image edge processing to detect the corresponding face image.

[0129] The Laplace operator used here is as follows:

[0130]

[0131] Taking the detected pixel point as the center, templates for detection in 16 directions such as 0°, 22.5°, 45°, 67.5°, 90°, 112.5°, 135°, 157.5°, 180°, 202.5°, 225°, 247.5°, 270°, 292.5°, 315°, and 337.5° are made. The estimation of the Laplace operator is as follows:

[0132]

[0133]

[0134] (4) Face alignment:

[0135] The face image detection module in the system preprocesses the collected face images, identifies the face images in the ID photos or real-time self-taken photos based on the extracted face image position and size information, and extracts the corresponding face images containing only the face for the ID photos or real-time self-taken photos.

[0136] The face image detection module in the system specifically uses the facial feature estimation algorithm to perform portrait recognition on the ID portrait photos and self-taken photos. That is, taking the frontal face facial feature points as the reference points, including the contour points of various parts of the face such as eyebrows, eyes, nose, cheekbones, mouth, and ears, detecting the corresponding contour points in the photo, and then using affine transformation to map the portrait in the photo to the reference point position to obtain a face image containing only the face.

[0137] (5) Feature extraction:

[0138] The face feature extraction module in the system extracts features from the face images containing only the face extracted during the face alignment process to extract discriminative face features.

[0139] The face feature extraction module in the system specifically inputs the face images containing only the face in the ID photos and the face images containing only the face in the self-taken photos into the fusion network model, and uses the dual-branch LeNet-5 convolutional neural network for feature extraction. The specific steps are as follows:

[0140] First, the edges of the input image with size M×N are padded with zero values, and the image size after padding is (M + 2*p)×(N + 2*p);

[0141] Next, take the matrix window X ij as all the pixels in the current convolutional sliding window, and perform a convolutional operation on the matrix window X ij with the weight matrix to obtain the eigenvalue Y ij ; The calculation formula here is:

[0142]

[0143] Among them, f(■) represents the activation function, W is the convolutional kernel weight matrix, and b is the bias value;

[0144] Next, after the convolution operation is completed, the obtained feature map is sampled by the max pooling method to obtain a series of feature matrices.

[0145] Finally, the feature matrices of the two pictures are integrated respectively to obtain two groups of feature vectors:

[0146] F left = 1×1×N li ;

[0147] F right = 1×1×N ri ;

[0148] Among them, N li , N ri respectively represent the number of neurons included in the i-th fully connected layer of the left and right branches.

[0149] (6) Key text extraction:

[0150] The key text extraction module in the system obtains the ID photo and performs image processing on the text information in the ID photo to extract key information such as the user's name and ID number.

[0151] The key text extraction module in the system specifically realizes key text extraction through the following steps:

[0152] First, the NLM algorithm is used to perform noise reduction processing on the ID photo. Similar regions are searched for on the image (i.e., the image of the ID photo) in units of image blocks, and then the average value is calculated. The largest window is the entire noise image (N×N), and a search window (L×L) is formed centered on the target pixel x and a neighborhood window (W×W) is formed centered on x and y, as Figure 4 shown.

[0153] When performing noise reduction processing, the neighborhood window centered on y slides in the search window, and weights are assigned by calculating the similarity degree between two neighborhood windows; after each pixel point in the search window is assigned a corresponding weight value, the pixel gray value after noise reduction is obtained.

[0154] Next, threshold values T 1 and T 2 are selected to perform binarization operation on the noise-reduced image. The formula is:

[0155]

[0156] Among them, V 0Indicates a non-text area, V 1 Indicates the currently uncertain area, V 2 Indicates the text area.

[0157] On this basis, for V 1 The eight-domain method is adopted for the area, that is, a set of 4×4 pixel points is processed each time, and then it is subjected to V 0 and V 2 division, and finally the binarization processing result is obtained.

[0158] Then, the skew correction and normalization processing are performed on the binarized image.

[0159] The processed image is a grayscale image containing character data, and the background pixels of the image and the pixels to be extracted and recognized have been preliminarily segmented. At the same time, the position of the text area has been basically determined, and the Hough transform is used to correct the skew of the image.

[0160] On this basis, the vertical projection value of each line of text in the text area is calculated using the vertical projection method to obtain the size of the average interval between characters. The text area is segmented using the periodic interval between characters in the text line, and the interpolation transformation method is used to normalize the text so that the text image has a unified size and ratio.

[0161] Then, character feature extraction is performed on the normalized image to obtain global stroke direction density features, local stroke direction density features, and peripheral stroke direction density features.

[0162] The direction features here can be defined by the following formula:

[0163]

[0164] where i ∈ [0,4), l i represents the distance size extending from this pixel point to the end of the character stroke in 8 directions.

[0165] Then, the model is trained according to the feature extraction results, and the text detection algorithm model based on deep learning is used to identify and extract the characters in the processed image.

[0166] Finally, the LCS algorithm is used to calculate the similarity between the name and ID number information extracted from the ID photo and the filled name and ID number information, calculate the matching score according to the similarity, output the matching result, and classify the matching result, such as successful matching and failed matching.

[0167] (7) Feature matching:

[0168] The similarity calculation and analysis module in the system compares the extracted face features with the corresponding features in the ID photo and calculates the similarity between the two.

[0169] The similarity calculation and analysis module in the system specifically uses the Contrastive Loss function to optimize the fusion network, thereby judging the similarity to determine whether the portrait in the ID is similar to the portrait in the selfie.

[0170] Among them, the expression of the Contrastive Loss function is:

[0171]

[0172] Among them, x 1 , x 2 is the input sample, and D w is the Euclidean distance between the samples (x 1 , x 2 ), and m is a hyperparameter.

[0173] (8) Similarity analysis:

[0174] The similarity calculation and analysis module in the system analyzes the similarity between two photos and generates a similarity score reflecting the matching degree between the two. If the similarity score is higher than the threshold, the verification passes; otherwise, it indicates a mismatch and the registration fails.

[0175] Specifically, the similarity distance metric is defined as:

[0176]

[0177] Among them, W i is the prior weight.

[0178] (9) Portrait determination result:

[0179] The similarity calculation and analysis module in the system determines whether the face image submitted by the user is the same person as the image in the ID photo according to the set threshold.

[0180] (10) Name comparison:

[0181] The registration information comparison module in the system compares the name filled in by the user with the name extracted from the ID photo in the key text extraction stage to ensure that the two are consistent.

[0182] (11) Comprehensive determination:

[0183] The comprehensive determination unit in the system combines the portrait determination result and the name comparison result for comprehensive determination. If both pass the verification, the user registration is successful; otherwise, the registration fails and the corresponding error message is given.

[0184] Based on the above examples, it can be seen that the alumni registration and authentication solution based on ID photo recognition given by the present invention has the following excellent effects in specific applications:

[0185] 1. High recognition efficiency and fast speed: By integrating advanced image recognition, computer vision, and machine learning technologies, this solution significantly improves the speed and accuracy of face recognition. In practical applications, compared with the traditional registration by filling in information, the registration speed is increased by 3 times.

[0186] 2. High security: The solution of the present invention introduces a live detection technology, which can effectively distinguish real human faces from non-live human faces such as photos, thus avoiding the risk of being fraudulently attacked. In addition, through a comprehensive determination mechanism (face determination and name comparison), the present invention further improves the accuracy and reliability of registration and authentication. During the data transmission and storage processes, the system ensures the security and privacy of user data through data encryption technology.

[0187] 3. Without excessive reliance on school resources: The solution of the present invention realizes the automation and intelligence of alumni registration and authentication, without relying on the resources of departments such as the school archives and information center. This not only reduces the school's manpower and material resources investment but also lowers the operating costs.

[0188] 4. Improve the work efficiency of the alumni association. The solution of the present invention reduces errors and omissions caused by human factors, saving valuable time and resources for the school.

[0189] 5. Provide more registration methods: The solution of the present invention can provide a convenient and efficient alumni registration method. Users only need to upload ID photos and take real-time selfies to complete the registration. This registration method is not only simple and easy to understand but also meets the registration needs of different users. Through this system, users can complete the registration anytime and anywhere without being restricted by time and place, improving the user's registration experience and satisfaction.

[0190] 6. Serve more universities and their alumni: The solution of the present invention has a wide application prospect and can be applied to the alumni registration and authentication systems of major universities. Through this solution, universities can more conveniently manage alumni information and strengthen the connection and communication with alumni.

[0191] In the above embodiments, the descriptions of each embodiment have their own emphases. For parts not detailed in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0192] Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working processes of the systems, devices, and modules described above can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0193] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memory, CD-ROM, optical memory, etc.) that contain computer-usable program code.

[0194] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram, and the combination of flows and / or blocks in the flowchart and / or block diagram, can be realized by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate means for realizing the functions specified in one Figure 1 flow or multiple flows and / or blocks Figure 1 block or multiple blocks.

[0195] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including instruction means, and the instruction means realizes the functions specified in one Figure 1 flow or multiple flows and / or blocks Figure 1 block or multiple blocks.

[0196] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for realizing the functions specified in one Figure 1 flow or multiple flows and / or blocks Figure 1 block or multiple blocks.

[0197] In a typical configuration, a computing device includes one or more processors (CPUs), an input / output interface, a network interface, and a memory.

[0198] The memory may include non-permanent memory in a computer-readable medium, in the form of random access memory (RAM) and / or non-volatile memory such as read-only memory (ROM) or flash memory (flash RAM). The memory is an example of a computer-readable medium.

[0199] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device.

[0200] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0201] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0202] The above-mentioned method of the present invention, or a specific system unit, or a part of the same, is a pure software architecture, which can be arranged on a physical medium, such as a hard disk, an optical disk, or any electronic device (such as a smart phone, a computer-readable storage medium) through program code. When the machine loads the program code and executes it (such as a smart phone loading and executing it), the machine becomes a device for implementing the present invention. The above-mentioned method and device of the present invention can also be transmitted in the form of program code through some transmission media, such as cables, optical fibers, or any transmission mode. When the program code is received, loaded and executed by a machine (such as a smart phone), the machine becomes a device for implementing the present invention.

[0203] The foregoing has shown and described the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments, and what is described in the above embodiments and the specification is only to illustrate the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements fall within the scope of the present invention claimed. The scope of the present invention claimed is defined by the appended claims and their equivalents.

Claims

1. A method for alumni registration and authentication based on ID photo recognition, characterized in that: include: (1) Obtaining the ID photo of the registrant and a real-time selfie photo containing the registrant's face, and pre-processing the ID photo and the real-time selfie photo to collect corresponding facial images containing only the face; (2) extracting distinctive facial features from the collected facial images, and performing similarity calculation and analysis on the facial images in the ID photo and the facial images in the real-time selfie photo based on the extracted facial features; (3) extracting key words from the obtained ID photo and comparing the extracted key words with the registration information entered by the registrant; (4) Determine the registration authentication result based on the facial image similarity calculation result in step (2) and the registration information comparison result in step (3).

2. The alumni registration authentication method based on ID photo recognition according to claim 1 is characterized in that: In the step (1), when preprocessing the ID photo, firstly, the position and size information of the face image are extracted from the ID photo; then, the face image in the ID photo is recognized based on the extracted position and size information of the face image, and the facial feature points of the front face image in the ID photo are used as reference points to detect the corresponding contour points in the ID photo, and then the face image in the ID photo is mapped to the reference point position using affine transformation, thereby obtaining a face image containing only the face.

3. The alumni registration authentication method based on ID photo recognition according to claim 1 is characterized in that: In the step (1), when preprocessing the real-time selfie photo, the real-time selfie photo is firstly subjected to a retake detection, and then the position and size information of the face image is extracted from the real-time selfie photo; then, the face image in the real-time selfie photo is recognized based on the extracted position and size information of the face image, and the facial feature points of the front face image in the real-time selfie photo are used as reference points to detect the corresponding contour points in the real-time selfie photo, and then the face image in the real-time selfie photo is matched to the reference point position by using affine transformation, thereby obtaining a face image containing only the face.

4. The alumni registration authentication method based on ID photo recognition according to claim 1 is characterized in that: In the step (2), the facial features with distinctiveness are extracted from the facial images based on a dual-branch LeNet-5 convolutional neural network.

5. The alumni registration authentication method based on ID photo recognition according to claim 4 is characterized in that: In the step (2), the contrastive loss function is used to optimize the fusion network.

6. An alumni registration and authentication system based on ID photo recognition, characterized in that: The alumni registration and authentication system includes an image acquisition and preprocessing unit, a feature extraction unit, an analysis and comparison unit, and a comprehensive determination unit. The image acquisition and preprocessing unit is configured to acquire the ID photo of the registrant and the real-time selfie photo containing the face of the registrant, and preprocess the acquired ID photo and the real-time selfie photo to acquire the corresponding face image containing only the face; The feature extraction unit is configured to perform data interaction with the image acquisition and preprocessing unit, and can extract distinctive face features from the acquired face images, and can extract key words from the acquired ID photo; The analysis and comparison unit is configured to perform data interaction with the feature extraction unit, and can perform similarity calculation and analysis on the face image in the ID photo and the face image in the real-time selfie photo according to the face features extracted by the feature extraction unit; and compare the key words extracted by the feature extraction unit with the registration information input by the registrant; The comprehensive determination unit is configured to perform data interaction with the analysis and comparison unit, and can determine the registration authentication result according to the facial image similarity calculation result and the registration information comparison result in the analysis and comparison unit.

7. The alumni registration and authentication system based on ID photo recognition according to claim 6 is characterized in that: The image acquisition and preprocessing unit comprises: An image acquisition module, the image acquisition module being configured to acquire a photo of the registrant's ID and a real-time selfie photo containing the registrant's face; A face image detection module, which is configured to interact with the image acquisition module data and can extract the position and size information of the corresponding face image from the collected ID photo or real-time selfie photo; A face image extraction module, wherein the face image extraction module is configured to interact with the face image detection module data, and can identify the face image in the ID photo or the real-time selfie photo based on the face image position and size information extracted by the face image detection module, using the facial feature points of the front face image in the ID photo or the real-time selfie photo as reference points, detecting the corresponding contour points in the ID photo or the real-time selfie photo, and then using affine transformation to correspond the face image in the ID photo or the real-time selfie photo to the reference point position, thereby obtaining a corresponding face image containing only the face.

8. The alumni registration and authentication system based on ID photo recognition according to claim 6 is characterized in that: The image acquisition and preprocessing unit also includes a liveness detection module, which is configured to interact with the image acquisition module data and can perform copy detection on the collected real-time selfie photos.

9. The alumni registration and authentication system based on ID photo recognition according to claim 6 is characterized in that: A fusion network model is constructed in the feature extraction unit, and a double-branch LeNet-5 convolutional neural network is used to extract distinctive facial features from facial images.

10. The alumni registration and authentication system based on ID photo recognition according to claim 9, characterized in that: The feature extraction unit also uses a contrastive loss function to optimize the fusion network.

Images