Candidate Identity Authentication Method, Device and System Based on Face Recognition Technology

Abstract

The present invention discloses a candidate identity verification method, device and system based on face recognition technology, which relates to the field of image processing technology. The method includes: acquiring multiple real-time face images of a candidate to be detected under multiple preset light source directions and multiple preset light source intensities; based on a reference face image, respectively performing region division on each real-time face image to obtain multiple face detection regions in each real-time face image; for each face detection region, respectively performing: determining the illumination consistency of the face detection region according to the gray values of each pixel point in the face detection region of each real-time face image, where the illumination consistency is used to characterize the degree of consistency of the influence of illumination on each pixel point in the face detection region; determining the reliability probability of the candidate to be detected according to the illumination consistency of each face detection region; and determining that the identity verification of the candidate to be detected is successful when the reliability probability of the candidate to be detected is greater than a preset reliability threshold.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and particularly to a method, device and system for verifying the identity of examinees based on face recognition technology. Background Art

[0002] Currently, face recognition technology has been widely applied in the examination security inspection system. The identity verification of examinees based on face recognition technology is to compare the facial image on the examinee's admission ticket with the facial image obtained by the face recognition device to determine whether the examinee is the person himself.

[0003] In the existing method, a multi-source face recognition technology based on active near-infrared light is adopted. By emitting infrared light with different angles and intensities, multiple facial images under different illumination directions and different illumination conditions are obtained, and the best facial image is selected from the multiple facial images to compare with the facial image on the examinee's admission ticket, so as to determine whether the examinee is the person himself.

[0004] However, since different parts of the face are affected by light to different degrees, some areas in the facial image have already appeared overexposed, while the illumination in some areas is still insufficient. The overall quality of the facial image obtained by the face recognition device is not high, resulting in a low accuracy of examinee identity verification. Summary of the Invention

[0005] Embodiments of the present invention provide a method, device and system for verifying the identity of examinees based on face recognition technology, which can improve the accuracy of examinee identity verification.

[0006] In the first aspect of the embodiments of the present invention, a method for verifying the identity of examinees based on face recognition technology is provided, including:

[0007] Obtaining multiple real-time facial images of the examinee to be detected under multiple preset light source directions and multiple preset light source intensities;

[0008] Based on a reference facial image, respectively dividing each real-time facial image into regions to obtain multiple facial detection regions in each real-time facial image, where the reference facial image is used to represent the correct facial image during examinee identity verification;

[0009] For each facial detection region, respectively perform: determining the illumination consistency of the facial detection region according to the gray values of each pixel point in the facial detection region of each real-time facial image, where the illumination consistency is used to represent the consistency degree of the influence of light on each pixel point in the facial detection region;

[0010] Determining the reliability probability of the examinee to be detected according to the illumination consistency of each facial detection region;

[0011] When the reliability probability of the candidate to be detected is greater than the preset reliability threshold, it is determined that the identity verification of the candidate to be detected is successful.

[0012] In the second aspect of the embodiments of the present invention, a candidate identity verification device based on face recognition technology is provided. The device includes: a memory and a program or instruction stored on the memory and executable on a processor. When the program or instruction is executed by the processor, it implements the candidate identity verification method based on face recognition technology provided in any aspect of the above embodiments of the present application.

[0013] In the third aspect of the embodiments of the present invention, a candidate identity verification system based on face recognition technology is provided, including:

[0014] An image acquisition module, configured to acquire multiple real-time face images of the candidate to be detected with multiple preset light source intensities in multiple preset light source directions;

[0015] A region division module, configured to respectively perform region division on each real-time face image based on a reference face image to obtain multiple face detection regions in each real-time face image, where the reference face image is used to represent the correct face image during candidate identity verification;

[0016] A region calculation module, configured to respectively execute for each face detection region: determine the illumination consistency of the face detection region according to the gray values of each pixel point in the face detection region of each real-time face image, where the illumination consistency is used to represent the degree of consistency of the influence of illumination on each pixel point in the face detection region;

[0017] A probability calculation module, configured to determine the reliability probability of the candidate to be detected according to the illumination consistency of each face detection region;

[0018] An identity verification module, configured to determine that the identity verification of the candidate to be detected is successful when the reliability probability of the candidate to be detected is greater than the preset reliability threshold.

[0019] In the candidate identity verification method based on face recognition technology provided by the embodiments of the present invention, based on a reference face image, a real-time face image is divided into regions to obtain multiple face detection regions. Then, for each face detection region, the lighting consistency of the face detection region is determined according to the gray values of each pixel point in the face detection region of each real-time face image. The lighting consistency is used to characterize the degree of consistency of the influence of light on each pixel point in the face detection region. Finally, the identity of the candidate to be detected is verified according to the lighting consistency of each face detection region. In this way, based on the reference face image, the present invention divides the real-time face image into multiple face detection regions, and evaluates the degree of influence of light on the pixel points in each face detection region, so as to accurately determine whether the identity verification of the candidate to be detected is successful. It can avoid the problem of low accuracy caused by directly judging from the overall face image and improve the accuracy of candidate identity verification. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained according to these drawings.

[0021] Figure 1 It is a schematic flowchart of the first candidate identity verification method based on face recognition technology provided by an embodiment of the present invention;

[0022] Figure 2 It is a schematic flowchart of the second candidate identity verification method based on face recognition technology provided by an embodiment of the present invention;

[0023] Figure 3 It is a schematic diagram of the result of the candidate identity verification system based on face recognition technology provided by an embodiment of the present invention;

[0024] Figure 4 It is a schematic diagram of the result of the candidate identity verification device based on face recognition technology provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0025] In order to further elaborate on the technical means and effects adopted by the present invention to achieve the intended invention purpose, the following will, in conjunction with the accompanying drawings and preferred embodiments, detail the specific implementation manners, structures, features and effects of a candidate identity verification method, device and system based on face recognition technology proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0026] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs.

[0027] It should be noted that the acquisition, storage, use, processing, etc. of data in the technical solution of the present invention all comply with the relevant regulations of laws and regulations.

[0028] It should be noted that in the embodiments of the present invention, some existing industry solutions such as certain software, components, models, etc. may be mentioned. They should be regarded as exemplary. The purpose is only to illustrate the feasibility in the implementation of the technical solution of the present invention, but it does not mean that the applicant has already or necessarily used this solution.

[0029] In the existing method, a multi-light source face recognition technology based on active near-infrared light is adopted. By emitting infrared light at different angles and intensities, multiple face images under different lighting directions and different lighting conditions are obtained. The best face image is selected from the multiple face images and compared with the face image on the candidate's admission ticket to determine whether the candidate is the person himself. However, the overall quality of the face images obtained by the face recognition device is not high, resulting in a low accuracy of candidate identity verification.

[0030] The purpose of the present invention is to provide a candidate identity verification method, device and system based on face recognition technology. In the candidate identity verification method based on face recognition technology provided by the embodiments of the present invention, based on a reference face image, the real-time face image is divided into regions to obtain multiple face detection regions. Then, for each face detection region, the lighting consistency of the face detection region is determined respectively according to the gray values of each pixel point in the face detection region of each real-time face image. The lighting consistency is used to characterize the degree of consistency of the influence of light on each pixel point in the face detection region. Finally, the identity of the candidate to be detected is verified according to the lighting consistency of each face detection region. In this way, based on the reference face image, the present invention divides the real-time face image into multiple face detection regions, and evaluates the degree of influence of light on the pixel points in each face detection region, so as to accurately determine whether the identity verification of the candidate to be detected is successful. It can avoid the problem of low accuracy caused by directly judging from the overall face image and improve the accuracy of candidate identity verification.

[0031] The following introduces specific embodiments of a candidate identity verification method, device, and system based on face recognition technology provided by the embodiments of the present invention.

[0032] First, the following introduces a candidate identity verification method based on face recognition technology provided by the embodiments of the present invention.

[0033] Figure 1 A flowchart of a candidate identity verification method based on face recognition technology is provided. This candidate identity verification method based on face recognition technology can be applied to a server and may include the following S101 to S105.

[0034] S101, obtain multiple real-time face images of the candidate to be detected at multiple preset light source directions and multiple preset light source intensities;

[0035] In this embodiment, the preset light source direction is used to represent the pre-set light source direction when collecting face images. Exemplarily, the preset light source directions may include directions such as the front, side, and top.

[0036] The preset light source intensity is used to represent the pre-set light intensity when collecting face images. Exemplarily, the preset light source intensities may include high-level light source intensity, medium-level light source intensity, and low-level light source intensity.

[0037] As an example, the server uses a face recognition device with adjustable light source direction and adjustable light source intensity to continuously or sequentially capture real-time face images of the candidate to be detected under different combinations of light source directions and light source intensities, and label the corresponding light source directions and light source intensities for each real-time face image of the candidate to be detected.

[0038] S102, based on the reference face image, respectively perform region division on each real-time face image to obtain multiple face detection regions in each real-time face image, where the reference face image is used to represent the correct face image during candidate identity verification;

[0039] In this embodiment, the reference face image is used to represent the pre-stored and verified correct face image of the candidate, which is usually collected during registration. Exemplarily, the reference face image may be the face image on the admission ticket.

[0040] As an example, the server uses a face recognition algorithm (such as Haar features, HOG+SVM, deep learning models, etc.) to identify and divide multiple face detection regions in the reference face image. For example, the face detection regions may include a right eye detection region, a left eye detection region, a left cheek detection region, and a right cheek detection region.

[0041] Then, according to the positions of the face detection regions in the reference face image, each real-time face image is respectively divided into regions to obtain the corresponding face detection regions in each real-time face image.

[0042] S103. For each face detection region, respectively execute: Determine the illumination consistency of the face detection region according to the gray values of the pixel points in the face detection region of each real-time face image, where the illumination consistency is used to characterize the degree of consistency of the influence of illumination on the pixel points in the face detection region.

[0043] In this embodiment, the illumination consistency is used to characterize the degree of consistency of the influence of illumination on the pixel points in the face detection region, that is, the uniformity of the gray values of the pixel points in the face detection region.

[0044] As an example, the server calculates the variance or standard deviation of the gray values of the pixel points in each real-time face image for each face detection region respectively, and calculates the mean value of the reciprocals of the variances or standard deviations of the gray values of the pixel points as the illumination consistency of the face detection region, so as to be used to evaluate the local illumination change in the face detection region.

[0045] Specifically, the larger the variance or standard deviation of the gray value of the pixel point, the greater the degree of fluctuation between the gray values of the pixel points in the face detection region. At this time, the illumination consistency of the face detection region is smaller.

[0046] S104. Determine the reliability probability of the candidate to be detected according to the illumination consistency of each face detection region.

[0047] In this embodiment, the server combines the illumination consistency of each face detection region and uses mechanisms such as weighted average, voting mechanism or machine learning model to evaluate and obtain the reliability probability of the candidate to be detected.

[0048] Specifically, in the case where the candidate to be detected is the person himself, the face detection region of the real-time face image should be consistent with the face detection region of the reference face image, that is, the finally divided face detection regions of the real-time face image are also the correct right eye detection region, left eye detection region, left cheek detection region and right cheek detection region.

[0049] And in the case where a face detection region contains pixel points on the same part of the human face, the degree of influence of illumination on them should be the same, that is, the illumination consistency is greater. Thus, the greater the illumination consistency of each face detection region, the higher the consistency between the face detection region of the real-time face image and the face detection region of the reference face image, that is, the greater the reliability probability of the candidate to be detected.

[0050] S105. When the reliability probability of the candidate to be detected is greater than the preset reliability threshold, it is determined that the identity verification of the candidate to be detected is successful.

[0051] In this embodiment, the server presets a preset reliability threshold in advance. Then, the determined reliability probability of the candidate to be detected is compared with this preset reliability threshold. When the reliability probability of the candidate to be detected is greater than the preset reliability threshold, it is determined that the identity verification of the candidate to be detected is successful, that is, the candidate to be detected is the candidate himself / herself; when the reliability probability of the candidate to be detected is less than or equal to the preset reliability threshold, it is determined that the identity verification of the candidate to be detected fails, that is, the candidate to be detected does not belong to the candidate himself / herself.

[0052] In the candidate identity verification method based on face recognition technology provided in this embodiment, based on the reference face image, the real-time face image is divided into regions to obtain multiple face detection regions. Then, for each face detection region, the illumination consistency of the face detection region is determined respectively according to the gray values of each pixel point in the face detection region of each real-time face image. The illumination consistency is used to characterize the degree of consistency of the influence of illumination on each pixel point in the face detection region. Finally, the identity of the candidate to be detected is verified according to the illumination consistency of each face detection region. In this way, based on the reference face image, the real-time face image is divided into multiple face detection regions in the present invention, and the influence degree of illumination on pixel points in each face detection region is evaluated, so as to accurately determine whether the identity verification of the candidate to be detected is successful. It can avoid the problem of low accuracy caused by directly judging from the overall face image and improve the accuracy of candidate identity verification.

[0053] As an alternative embodiment, before S102, the candidate identity verification method based on face recognition technology may further include:

[0054] Calculate the similarity between each real-time face image and the reference face image respectively to obtain the image similarity between each real-time face image and the reference face image;

[0055] When the maximum similarity among the image similarities is greater than the first similarity threshold, it is determined that the identity verification of the candidate to be detected is successful;

[0056] When the maximum similarity among the image similarities is less than the second similarity threshold, it is determined that the identity verification of the candidate to be detected fails, and the second similarity threshold is less than the first similarity threshold;

[0057] S102 may specifically include:

[0058] When the maximum similarity among the image similarities is not greater than the first similarity threshold and not less than the second similarity threshold, based on the reference face image, each real-time face image is respectively divided into regions to obtain multiple face detection regions in each real-time face image.

[0059] In this embodiment, the first similarity threshold is used to determine whether the candidate to be detected belongs to the candidate himself / herself, and the second similarity threshold is used to determine whether the candidate to be detected does not belong to the candidate himself / herself. The second similarity threshold is less than the first similarity threshold.

[0060] As an example, the server uses a face recognition algorithm (such as a deep learning model) to extract the first feature vector of the reference face image, and at the same time uses the same face recognition algorithm to extract the second feature vector of each real-time face image. Then, a target similarity calculation method (such as cosine similarity, Euclidean distance, and Manhattan distance, etc.) is used to compare the similarity between the feature vector of the real-time face image and the feature vector of the reference face image to obtain the image similarity between each real-time face image and the reference face image.

[0061] Then, find the maximum similarity among the image similarities and compare this maximum similarity with the first similarity threshold and the second similarity threshold respectively. When the maximum similarity is greater than the first similarity threshold, it is directly determined that the candidate to be detected has successfully passed the identity verification, that is, the candidate to be detected belongs to the candidate himself / herself; when the maximum similarity is less than the second similarity threshold, it is directly determined that the candidate to be detected has failed the identity verification, that is, the candidate to be detected does not belong to the candidate himself / herself; and when the maximum similarity is not greater than the first similarity threshold and not less than the second similarity threshold, at this time, it is impossible to directly determine whether the candidate to be detected belongs to the candidate himself / herself, so each real-time face image is respectively divided into regions, and the identity verification of the candidate to be detected is carried out according to the above-mentioned candidate identity verification method based on face recognition technology.

[0062] Through this embodiment, the first similarity threshold and the second similarity threshold are set, and then the maximum similarity among the image similarities between each real-time face image and the reference face image is compared with the first similarity threshold and the second similarity threshold respectively. In this way, it is possible to directly and simply complete the identity verification of the candidate to be detected in the case where the candidate to be detected is clearly the candidate himself / herself and the candidate to be detected is clearly not the candidate himself / herself, thereby improving the identity verification efficiency of the candidate to be detected.

[0063] As an alternative embodiment, S102 may specifically include:

[0064] Based on the size of the face region in the reference face image, each real-time face image is respectively adjusted in size so that the size of the face region in each real-time face image is equal to the size of the face region in the reference face image, and each real-time face image after size adjustment is obtained;

[0065] The reference face image is divided into multiple face detection regions by using an edge detection algorithm;

[0066] Based on the regional positions of the face detection regions in the reference face image, each real-time face image after size adjustment is respectively divided into regions, and multiple face detection regions in each real-time face image are obtained.

[0067] In this embodiment, the server uses a face detection algorithm (such as Haar features, HOG+SVM, deep learning model, etc.) to respectively identify the face region in the reference face image and the real-time face image. For each real-time face image, calculate the width and height of its face region, and then calculate the scaling ratio with the reference size of the face region in the reference face image. Then use an image scaling algorithm (such as bilinear interpolation, nearest neighbor interpolation, etc.) to scale each real-time face image so that the size of its face region is equal to the reference size of the face region in the reference face image.

[0068] Then, use an edge detection algorithm (such as Canny edge detector, Sobel operator, Laplacian operator, etc.) to perform edge detection on the reference face image to obtain the edge information of the face region in the reference face image. Then, according to the edge information, the reference face image is divided into multiple regions according to facial features.

[0069] Finally, according to the position information (such as center coordinates, width, height, etc.) of each face detection region in the reference face image, establish the mapping relationship between these face detection regions and the corresponding regions in the real-time face image. Thus, according to the mapping relationship, the regions corresponding to the face detection regions in the reference face image are divided in the real-time face image after size adjustment.

[0070] Through this embodiment, each real-time face image is first adjusted in size so that the size of the face region in each real-time face image is equal to the size of the face region in the reference face image. Then, according to the multiple face detection regions in the reference face image, each real-time face image is respectively divided into regions. In this way, the regional division of the real-time face image can be accurately completed, which helps to verify the identity of the examinee according to the face detection region of the real-time face image.

[0071] As an alternative embodiment, as Figure 2As shown in the figure, S103 may specifically include: for each face detection area, the following S201 to S205 are respectively executed.

[0072] S201. According to the gray values of each pixel point in the face detection area in each real-time face image, determine the illumination influence degree of each pixel point in each preset light source direction respectively. The illumination influence degree is used to characterize the influence degree of the pixel point affected by the light intensity.

[0073] S202. According to the illumination influence degree of each pixel point in each preset light source direction, determine the local illumination consistency of the face detection area in each preset light source direction respectively.

[0074] S203. Based on the average gray value of the pixel points under the maximum light source intensity in each preset light source direction of the face detection area, set the weights of each preset light source direction respectively.

[0075] S204. Based on the weights of each preset light source direction, cluster each preset light source direction to obtain at least one light source direction cluster.

[0076] S205. According to the local illumination consistency of the face detection area in each preset light source direction and the light source direction cluster, determine the illumination consistency of the face detection area.

[0077] In this embodiment, the illumination influence degree is used to characterize the influence degree of the pixel point affected by the light intensity in the preset light source direction.

[0078] The local illumination consistency is used to characterize the local illumination consistency of the face detection area only in a certain preset light source direction.

[0079] As an example, for each face detection area, the server respectively executes the following steps:

[0080] First, for each pixel point in the face detection area, obtain the gray values of the real-time face images corresponding to different preset light source intensities in the same preset light source direction. According to the change situation of the gray values of the pixel point in different preset light source intensities in the preset light source direction, the illumination influence degree of the pixel point in the preset light source direction can be determined. Specifically, the greater the change of the gray value of the pixel point in different preset light source intensities in the preset light source direction, the greater the illumination influence degree; the smaller the change of the gray value of the pixel point in different preset light source intensities in the preset light source direction, the smaller the illumination influence degree. In this way, the illumination influence degree of each pixel point in each preset light source direction in the face detection area can be obtained through the above steps.

[0081] Then, for each preset light source direction, calculate the variance or standard deviation of the illumination influence degrees of all pixel points within the face detection region in this preset light source direction, and use the reciprocal of this variance or standard deviation as the local illumination consistency of the face detection region in this preset light source direction. Specifically, the smaller the difference between the illumination influence degrees of each pixel point within the face detection region in this preset light source direction, the greater the local illumination consistency of the face detection region in this preset light source direction; the greater the difference between the illumination influence degrees of each pixel point within the face detection region in this preset light source direction, the smaller the local illumination consistency of the face detection region in this preset light source direction.

[0082] Then, calculate the gray-scale mean value of the pixel points in the maximum light source intensity of the face detection region in each preset light source direction, and set the weight of the face detection region in each preset light source direction through the following formula 1:

[0083] Formula 1

[0084] In formula 1, is used to represent the weight of the a-th face detection region in the b-th preset light source direction, is used to represent the gray-scale mean value of the pixel points in the maximum light source intensity of the a-th face detection region in the b-th preset light source direction, is used to represent the weight normalization function.

[0085] Then, the server selects a suitable clustering algorithm according to actual requirements, such as K-means, hierarchical clustering, etc. Using the selected clustering algorithm, cluster each preset light source direction according to the weights of the face detection region in each preset light source direction to obtain at least one light source direction cluster, and each cluster contains multiple preset light source directions with similar weights.

[0086] Finally, for each light source direction cluster, evaluate the local illumination consistency of the face detection region under each preset light source direction within it. Specifically, statistical analysis can be performed on the local illumination consistency of the face detection region under each preset light source direction within the light source direction cluster to obtain the local illumination consistency corresponding to the light source direction cluster. Then, according to the local illumination consistency corresponding to each light source direction cluster, use mechanisms such as weighted average, voting mechanism, or machine learning model to evaluate and obtain the illumination consistency of the face detection region.

[0087] Through this embodiment, based on the gray values of the pixel points in the real-time face image of each preset light source intensity in each preset light source direction in the face detection area, the light consistency of the pixel points in the face detection area affected by light is comprehensively determined. In this way, through comprehensive analysis from multiple preset light source directions and multiple preset light source intensities, the light consistency of the pixel points in the face detection area affected by light can be accurately evaluated, thereby improving the accuracy of candidate identity verification.

[0088] As an alternative embodiment, S201 may specifically include:

[0089] For each pixel point, the following steps are respectively executed in each preset light source direction:

[0090] Subtract the gray value of the pixel point at the d-th preset light source intensity in the preset light source direction from the gray value of the pixel point at the (d - 1)-th preset light source intensity in the preset light source direction to obtain the d-th gray value difference, where d is a positive integer;

[0091] Divide the d-th gray value difference by the target gray value to obtain the local light influence degree of the pixel point at the d-th preset light source intensity in the preset light source direction, where the target gray value is the gray value of the pixel point at the maximum preset light source intensity in the preset light source direction;

[0092] Calculate the mean value of the local light influence degrees of the pixel point at each preset light source intensity in the preset light source direction to obtain the light influence degree of the pixel point in the preset light source direction.

[0093] In this embodiment, the local light influence degree of the pixel point in the preset light source direction can be specifically determined by the following formula 2:

[0094] Formula 2

[0095] In formula 2, is used to represent the d-th local light influence degree of the i-th pixel point in the a-th face detection area in the b-th preset light source direction. is used to represent the gray value of the i-th pixel point in the a-th face detection area at the d-th preset light source intensity in the b-th preset light source direction, is used to represent the gray value of the i-th pixel point in the a-th face detection area at the (d - 1)-th preset light source intensity in the b-th preset light source direction. is used to represent the gray value of the i-th pixel point in the a-th face detection area at the maximum preset light source intensity in the b-th preset light source direction, and its value is not 0.

[0096] Among them, as the light source intensity increases, the gray value of the pixel point also increases. The larger the value is, it indicates that the influence of the i-th pixel in the a-th face detection area by the b-th preset light source direction is greater, and the illumination influence degree of the i-th pixel in the a-th face detection area under the b-th preset light source direction is greater.

[0097] After obtaining the local illumination influence degrees of the pixel under the preset light source directions, the average value of each local illumination influence degree is calculated to obtain the illumination influence degree of the pixel under the preset light source direction.

[0098] Through this embodiment, according to the differences between the gray values of the pixel under each preset light source intensity in the preset light source direction, the illumination influence degree of the pixel under the preset light source direction can be accurately evaluated. In this way, by accurately calculating the illumination influence degrees of each pixel under each preset light source direction, the accuracy of candidate identity verification can be improved.

[0099] As an optional embodiment, S202 may specifically include:

[0100] For the face detection area, the following steps are respectively executed under each preset light source direction:

[0101] Based on the illumination influence degrees of each pixel under the preset light source direction, each pixel is clustered to obtain at least one pixel cluster;

[0102] Divide the number of pixels in the largest pixel cluster by the total number of pixels in the face detection area to obtain the local illumination consistency of the face detection area under the preset light source direction.

[0103] In this embodiment, for the face detection area, the server respectively executes the following steps under each preset light source direction:

[0104] First, the server clusters each pixel based on the illumination influence degrees of each pixel in the face detection area under the preset light source direction by using a density-based clustering algorithm to obtain at least one pixel cluster, and each pixel cluster contains multiple pixels with similar illumination influence degrees.

[0105] Then, the local illumination consistency of the face detection area under the preset light source direction is determined by the following formula 3:

[0106] Formula 3

[0107] In formula 3, is used to represent the local illumination consistency of the a-th face detection area under the b-th preset light source direction, is used to represent the number of pixels in the largest pixel cluster of the a-th face detection area under the b-th preset light source direction, Used to represent the total number of pixel points in the a-th face detection region.

[0108] Among them, The larger it is, the more similar the degree of influence of the pixel points in the a-th face detection region under the b-th preset light source direction is, that is, the greater the possibility that the pixel points in the a-th face detection region belong to the same part of the human face, and the higher the consistency between the face detection region of the real-time face image and the face detection region of the reference face image, so the reliability probability of the candidate to be detected is greater.

[0109] Through this embodiment, based on the illumination influence degree of each pixel point in the face detection region under the preset light source direction, the local illumination consistency of the face detection region under the preset light source direction can be accurately evaluated. In this way, by accurately calculating the local illumination consistency of the face detection region under each preset light source direction, the accuracy of candidate identity verification can be improved.

[0110] As an optional embodiment, S205 may specifically include:

[0111] Perform variance calculations on the local illumination consistency of the face detection region under the preset light source directions in each light source direction cluster respectively, to obtain the illumination consistency volatility of the face detection region in each light source direction cluster;

[0112] Determine the light source direction continuity corresponding to the face detection region and each light source direction cluster according to the number of preset light source directions of the face detection region in each light source direction cluster;

[0113] Determine the illumination consistency of the face detection region according to each local illumination consistency, each illumination consistency volatility, and each light source direction continuity.

[0114] In this embodiment, the server performs variance calculations on the local illumination consistency of the face detection region under each preset light source direction in the light source direction cluster, to obtain the illumination consistency volatility of the face detection region in the light source direction cluster.

[0115] Then, subtract 2 from the number of preset light source directions of the face detection region in the light source direction cluster, to obtain the light source direction continuity corresponding to the face detection region and the light source direction cluster.

[0116] Finally, according to each local illumination consistency, each illumination consistency volatility, and each light source direction continuity, determine the illumination consistency of the face detection region through the following formula 4:

[0117] Formula 4

[0118] In formula 4, Used to represent the illumination consistency of the a-th face detection region, used to characterize the local illumination consistency of the a-th face detection region under the b-th preset light source direction used to represent the weight of the a-th face detection region under the b-th preset light source direction, and n is used to represent the total number of preset light source directions used to represent the mean value of the continuity of each light source direction of the a-th face detection region, and its value is not zero used to represent the mean value of the fluctuation of each illumination consistency of the a-th face detection region, and exp is used to represent the exponential operation

[0119] Wherein The smaller the value of, the more it conforms to the characteristic that when the pixel points of the a-th face detection region belong to the same part of the human face, multiple connected preset light source directions will be grouped into the same type of cluster, and then the illumination consistency of the a-th face detection region is greater The smaller the value of, the more it conforms to the characteristic that when the pixel points of the a-th face detection region belong to the same part of the human face, the local illumination consistencies of all preset light source directions in the light source direction cluster are closer, and then the illumination consistency of the a-th face detection region is greater

[0120] Through this embodiment, according to the illumination consistency fluctuation of the face detection region in each light source direction cluster, the light source direction continuity corresponding to the face detection region and each light source direction cluster, and the local illumination consistency of the face detection region under the preset light source direction, the illumination consistency of the face detection region can be accurately evaluated. In this way, by accurately calculating the illumination consistency of the face detection region, the accuracy of candidate identity verification can be improved

[0121] As an optional embodiment, S104 may specifically include

[0122] Determine the control detection region of each face detection region according to the number of pixel points in each face detection region, and the control detection region is the region where the difference in the number of pixel points from the face detection region is less than the pixel point difference threshold

[0123] Determine the reliability probability of the candidate to be detected according to the illumination consistency of each face detection region and the illumination consistency of the corresponding control detection region

[0124] In this embodiment, since the human face is a left-right symmetric image. Therefore, if the candidate to be detected belongs to the candidate himself, there will be some face detection regions with similar illumination consistencies. For example, the illumination consistencies of the left eye detection region and the right eye detection region are similar

[0125] Therefore, the server first obtains the number of pixel points in each face detection area. Then, for each face detection area, other face detection areas with a difference in the number of pixel points less than 5 from the face detection area are determined as the control detection areas of the face detection area. For example, in the case where there are multiple other face detection areas that meet this condition, the one with the smallest difference is determined as the control detection area of the face detection area. In this way, the control detection areas of each face detection area can be obtained.

[0126] Then, according to the lighting consistency of each face detection area and the lighting consistency of the corresponding control detection area, the reliability probability of the candidate to be detected is determined through the following formula 5:

[0127] Formula 5

[0128] In formula 5, S is used to represent the reliability probability of the candidate to be detected, is used to represent the lighting consistency of the b-th face detection area, is used to represent the lighting consistency of the control detection area of the b-th face detection area, and M is used to represent the number of face detection areas with control detection areas. is used to represent the total number of pixel points in the a-th face detection area, is used to represent the lighting consistency of the a-th face detection area. exp is used to represent the exponential operation, is used to represent the weight normalization function, and j is used to represent the total number of face detection areas.

[0129] Among them, The larger the value of, the greater the lighting consistency of multiple face detection areas of the candidate to be detected, which means the greater the possibility that the candidate to be detected belongs to the candidate himself / herself, that is, the greater the reliability probability of the candidate to be detected; The smaller the value of, the more the face detection area of the candidate to be detected conforms to the characteristics of left-right symmetry of the human face with its corresponding control detection area, which means the greater the possibility that the candidate to be detected belongs to the candidate himself / herself, that is, the greater the reliability probability of the candidate to be detected.

[0130] Through this embodiment, according to the lighting consistency of each face detection area and the lighting consistency of the corresponding control detection area, the reliability probability of the candidate to be detected can be accurately evaluated. In this way, by accurately calculating the reliability probability of the candidate to be detected, the accuracy of candidate identity verification can be improved.

[0131] According to the candidate identity verification method based on face recognition technology. Correspondingly, the present invention also provides a specific embodiment of a candidate identity verification system based on face recognition technology.

[0132] Figure 3The schematic structural diagram of the candidate identity verification system based on face recognition technology provided by the embodiments of the present application is shown. The candidate identity verification system 300 based on face recognition technology may include an image acquisition module 310, a region division module 320, a region calculation module 330, a probability calculation module 340, and an identity verification module 350.

[0133] The image acquisition module 310 is configured to acquire multiple real-time face images of the candidate to be detected under multiple preset light source directions and multiple preset light source intensities.

[0134] The region division module 320 is configured to perform region division on each real-time face image respectively based on a reference face image to obtain multiple face detection regions in each real-time face image. The reference face image is used to represent the correct face image during candidate identity verification.

[0135] The region calculation module 330 is configured to, for each face detection region, perform the following respectively: determine the lighting consistency of the face detection region according to the gray values of each pixel point in the face detection region of each real-time face image. The lighting consistency is used to represent the degree of consistency of the influence of light on each pixel point in the face detection region.

[0136] The probability calculation module 340 is configured to determine the reliability probability of the candidate to be detected according to the lighting consistency of each face detection region.

[0137] The identity verification module 350 is configured to determine that the identity verification of the candidate to be detected is successful when the reliability probability of the candidate to be detected is greater than a preset reliability threshold.

[0138] In the candidate identity verification system based on face recognition technology provided by this embodiment, based on the reference face image, the real-time face image is divided into multiple face detection regions. Then, for each face detection region, the lighting consistency of the face detection region is determined respectively according to the gray values of each pixel point in the face detection region of each real-time face image. The lighting consistency is used to represent the degree of consistency of the influence of light on each pixel point in the face detection region. Finally, the identity of the candidate to be detected is verified according to the lighting consistency of each face detection region. In this way, based on the reference face image, the present invention divides the real-time face image into multiple face detection regions, and evaluates from the degree of influence of light on the pixel points in each face detection region, so as to accurately determine whether the identity verification of the candidate to be detected is successful. It can avoid the problem of low accuracy caused by directly judging from the overall face image and improve the accuracy of candidate identity verification.

[0139] According to the candidate identity verification method based on face recognition technology. Correspondingly, the present invention also provides a specific embodiment of a candidate identity verification device based on face recognition technology.

[0140] Figure 4 The figure shows a schematic diagram of the hardware structure of the candidate identity verification device based on face recognition technology provided by an embodiment of the present invention.

[0141] The candidate identity verification device based on face recognition technology may include a processor 401 and a memory 402 storing computer program instructions.

[0142] Specifically, the above-mentioned processor 401 may include a central processing unit, or a specific integrated circuit, or one or more integrated circuits configured to implement the embodiments of the present application.

[0143] The memory 402 may include a mass storage for data or instructions. By way of example and not limitation, the memory 402 may include a hard disk drive, a floppy disk drive, a flash memory, an optical disc, a magneto-optical disc, a magnetic tape, or a universal serial bus drive, or a combination of two or more of these. In a suitable case, the memory 402 may include a removable or non-removable (or fixed) medium. In a suitable case, the memory 402 may be inside or outside the integrated gateway disaster recovery device. In a specific embodiment, the memory 402 is a non-volatile solid-state memory.

[0144] The memory 402 may include a read-only memory, a random access memory, a magnetic disk storage medium device, an optical storage medium device, a flash memory device, an electrical, optical, or other physical / tangible memory storage device. Thus, generally, the memory 402 includes one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the method according to one aspect of the present disclosure.

[0145] The processor 401 reads and executes the computer program instructions stored in the memory 402 to implement any one of the candidate identity verification methods based on face recognition technology in the above embodiments.

[0146] In one example, the digital twin model construction device in a complex scenario may further include a communication interface 403 and a bus 410. Among them, as Figure 4 shown, the processor 401, the memory 402, and the communication interface 403 are connected through the bus 410 and complete communication with each other.

[0147] The communication interface 403 is mainly used to implement communication between each module, device, unit, and / or device in the embodiments of the present application.

[0148] The bus 410 includes hardware, software, or both, and couples the components of the candidate authentication device based on face recognition technology to each other. By way of example and not limitation, the bus may include an Accelerated Graphics Port or other graphics bus, an Enhanced Industry Standard Architecture bus, a Front Side Bus, a HyperTransport interconnect, an Industry Standard Architecture bus, an InfiniBand interconnect, a Low Pin Count bus, a Memory bus, a MicroChannel Architecture bus, a Peripheral Component Interconnect bus, a Serial Advanced Technology Attachment bus, a Video Electronics Standards Association Local bus, or other suitable bus or a combination of two or more of these. Where appropriate, the bus 410 may include one or more buses. Although the embodiments of the present application describe and illustrate specific buses, the present application contemplates any suitable bus or interconnect.

[0149] It should be clear that the present invention is not limited to the specific configurations and processes described above and illustrated in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and illustrated as examples. However, the method process of the present invention is not limited to the specific steps described and illustrated, and those skilled in the art can make various changes, modifications, and additions, or change the order between steps after understanding the spirit of the present invention.

[0150] It should also be noted that the exemplary embodiments mentioned in the present invention describe some methods or systems based on a series of steps or devices. However, the present invention is not limited to the order of the above steps, that is, the steps can be executed in the order mentioned in the embodiments, or different from the order in the embodiments, or several steps can be executed simultaneously.

[0151] As described above, only the specific embodiments of the present invention are provided. Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the systems, modules, and units described above can refer to the corresponding processes in the foregoing method embodiments, and will not be repeated here. It should be understood that the protection scope of the present invention is not limited thereto, and any person skilled in the art within the technical scope disclosed by the present invention can easily think of various equivalent modifications or substitutions, and these modifications or substitutions should all be covered within the protection scope of the present invention.

Claims

1. A method for verifying the identity of examinees based on face recognition technology, characterized in that: The method comprises: Acquire multiple real-time facial images of the examinee to be tested under multiple preset light source directions and multiple preset light source intensities; Based on the reference facial image, each of the real-time facial images is divided into regions to obtain a plurality of facial detection regions in each of the real-time facial images, wherein the reference facial image is used to represent the correct facial image for the identity verification of the examinee; For each of the face detection areas, respectively performing: determining the illumination consistency of the face detection area according to the grayscale value of each pixel point in the face detection area of ​​each of the real-time face images, wherein the illumination consistency is used to characterize the consistency degree of each pixel point in the face detection area affected by illumination; Determining the reliability probability of the examinee to be detected according to the illumination consistency of each of the face detection areas; In the case where the reliability probability of the candidate to be tested is greater than a preset reliability threshold, determining that the identity authentication of the candidate to be tested is successful; For each of the face detection areas, the following steps are performed respectively: According to the grayscale value of each pixel point in the face detection area in each of the real-time face images, respectively determine the illumination influence degree of each pixel point under each of the preset light source directions, wherein the illumination influence degree is used to characterize the degree to which the pixel point is affected by the light intensity; Determining the local illumination consistency of the face detection area in each of the preset light source directions according to the illumination influence of each of the pixel points in each of the preset light source directions; Based on the mean grayscale value of the pixels in the face detection area under the maximum light intensity in each of the preset light source directions, respectively setting the weight of each of the preset light source directions; Based on the weight of each of the preset light source directions, clustering the preset light source directions to obtain at least one light source direction cluster; The illumination consistency of the face detection area is determined according to the local illumination consistency of the face detection area under each of the preset light source directions and the light source direction cluster.

2. The examinee identity verification method based on face recognition technology according to claim 1 is characterized in that: Before dividing each of the real-time facial images into regions based on the reference facial image to obtain a plurality of facial detection regions in each of the real-time facial images, the method further includes: Calculating the similarity between each of the real-time facial images and the reference facial image to obtain the image similarity between each of the real-time facial images and the reference facial image; When the maximum similarity among the image similarities is greater than the first similarity threshold, it is determined that the identity authentication of the candidate to be detected is successful; In the case where the maximum similarity among the image similarities is less than the second similarity threshold, it is determined that the identity authentication of the candidate to be detected has failed, and the second similarity threshold is less than the first similarity threshold; The step of dividing each of the real-time facial images into regions based on the reference facial image to obtain a plurality of facial detection regions in each of the real-time facial images comprises: When the maximum similarity among the image similarities is not greater than the first similarity threshold and not less than the second similarity threshold, each of the real-time facial images is divided into regions based on a reference facial image to obtain a plurality of facial detection regions in each of the real-time facial images.

3. The examinee identity verification method based on face recognition technology according to claim 1 is characterized in that: The step of dividing each of the real-time facial images into regions based on the reference facial image to obtain a plurality of facial detection regions in each of the real-time facial images comprises: Based on the size of the facial region in the reference facial image, resize each of the real-time facial images so that the size of the facial region in each of the real-time facial images is equal to the size of the facial region in the reference facial image, thereby obtaining each of the real-time facial images after resizing; By using an edge detection algorithm, the reference facial image is divided into regions to obtain a plurality of facial detection regions in the reference facial image; Based on the area position of each face detection area in the reference face image, each of the real-time face images after the size adjustment is respectively divided into areas to obtain a plurality of face detection areas in each of the real-time face images.

4. The examinee identity verification method based on face recognition technology according to claim 1 is characterized in that: The step of determining the illumination influence of each pixel in each of the preset light source directions according to the grayscale value of each pixel in the face detection area in each of the real-time face images comprises: For each pixel point, the following steps are performed respectively under each preset light source direction: Subtract the grayscale value of the pixel point at the d-th preset light source intensity in the preset light source direction from the grayscale value of the pixel point at the d-1th preset light source intensity in the preset light source direction to obtain the d-th grayscale value difference, where d is a positive integer; Dividing the d-th grayscale value difference by the target grayscale value, the local illumination influence of the pixel point under the d-th preset light source intensity in the preset light source direction is obtained, and the target grayscale value is the grayscale value of the pixel point under the maximum preset light source intensity in the preset light source direction; The local illumination influence of the pixel point under each of the preset light source intensities in the preset light source direction is averaged to obtain the illumination influence of the pixel point under the preset light source direction.

5. The examinee identity verification method based on face recognition technology according to claim 1 is characterized in that: The determining, according to the illumination influence of each pixel point in each of the preset light source directions, respectively the local illumination consistency of the face detection area in each of the preset light source directions, comprises: For the face detection area, the following steps are performed respectively under each of the preset light source directions: Based on the illumination influence of each pixel point in the preset light source direction, clustering each pixel point to obtain at least one pixel point cluster; The number of pixels in the largest pixel cluster is divided by the total number of pixels in the face detection area to obtain the local illumination consistency of the face detection area in the preset light source direction.

6. The examinee identity verification method based on face recognition technology according to claim 1 is characterized in that: The determining the illumination consistency of the face detection area according to the local illumination consistency of the face detection area under each of the preset light source directions and the light source direction cluster comprises: Performing variance calculations on the local illumination consistency of the face detection area under the preset light source directions in each of the light source direction clusters to obtain illumination consistency fluctuations of the face detection area in each of the light source direction clusters; Determining the continuity of the light source directions corresponding to the face detection area and each of the light source direction clusters according to the number of the preset light source directions of the face detection area in each of the light source direction clusters; The illumination consistency of the face detection area is determined according to the local illumination consistency, the illumination consistency fluctuation and the light source direction continuity.

7. The examinee identity verification method based on face recognition technology according to any one of claims 1 to 3, characterized in that: The step of determining the reliability probability of the examinee to be detected according to the illumination consistency of each of the face detection areas includes: Determine, according to the number of pixels in each face detection area, a control detection area for each face detection area, wherein the control detection area is an area whose difference in the number of pixels between the face detection area and the face detection area is less than a pixel difference threshold; The reliability probability of the examinee to be detected is determined based on the illumination consistency of each of the face detection areas and the illumination consistency of the corresponding control detection area.

8. A candidate identity verification device based on face recognition technology, characterized in that: The device comprises: a processor and a memory storing computer program instructions; When the processor executes the computer program instructions, it implements the candidate identity authentication method based on face recognition technology as described in any one of claims 1-7.

9. A candidate identity verification system based on face recognition technology, characterized in that: The system comprises: An image acquisition module is used to acquire multiple real-time facial images of the examinee under multiple preset light source directions and multiple preset light source intensities; A region division module, for performing region division on each of the real-time facial images based on a reference facial image, to obtain a plurality of facial detection regions in each of the real-time facial images, wherein the reference facial image is used to represent a correct facial image for identity verification of the examinee; The area calculation module is used to respectively perform, for each of the face detection areas: determining the illumination consistency of the face detection area according to the grayscale value of each pixel in the face detection area of ​​each of the real-time facial images, wherein the illumination consistency is used to characterize the consistency degree of each pixel in the face detection area affected by illumination; and respectively perform the following steps for each of the face detection areas: According to the grayscale value of each pixel point in the face detection area in each of the real-time face images, respectively determine the illumination influence degree of each pixel point under each of the preset light source directions, wherein the illumination influence degree is used to characterize the degree to which the pixel point is affected by the light intensity; Determining the local illumination consistency of the face detection area in each of the preset light source directions according to the illumination influence of each of the pixel points in each of the preset light source directions; Based on the mean grayscale value of the pixels in the face detection area under the maximum light intensity in each of the preset light source directions, respectively setting the weight of each of the preset light source directions; Based on the weight of each of the preset light source directions, clustering the preset light source directions to obtain at least one light source direction cluster; Determining the illumination consistency of the face detection area according to the local illumination consistency of the face detection area under each of the preset light source directions and the light source direction cluster; A probability calculation module, used to determine the reliability probability of the examinee to be detected according to the illumination consistency of each face detection area; The identity authentication module is used to determine that the identity authentication of the candidate to be detected is successful when the reliability probability of the candidate to be detected is greater than a preset reliability threshold.

Images