Face recognition processing method and device

Abstract

The present invention provides a face recognition processing method and device, which relates to the field of recognition technology and can be used in the financial field or other technical fields. The method includes: acquiring a face image of a user, performing image recognition on the face image, and obtaining the scene environment parameters and user characteristic parameters of the shooting of the face image; determining the original scores of each model provided by each face recognition multi-model fusion provider according to the scene environment parameters and the user characteristic parameters; rating each face recognition multi-model fusion provider according to the original score of each model, determining the target face recognition multi-model fusion provider that executes the face recognition multi-model fusion decision according to the rating results, and performing face recognition on the user based on the model provided by the target face recognition multi-model fusion provider. The device executes the above method. The method and device provided by the embodiment of the present invention can ensure the accuracy of face recognition results.

Description

Technical Field

[0001] The present invention relates to the field of recognition technology, and in particular to a face recognition processing method and device. Background Art

[0002] Facial recognition technology has been widely used in commercial banks' product services, customer marketing, operations management, and risk management, including facial payment, at-the-counter ID verification, and facial access control / attendance. Currently, the financial sector relies on a single facial algorithm model for recognition, and the limitations of decision-making based on this single algorithm model are gradually becoming apparent. These limitations are primarily reflected in the following: existing facial recognition technology is based on iterative upgrades of a single algorithm model. In complex scenarios, the complexity of the environment and customer base often leads to a decrease in the algorithm's recognition accuracy. Furthermore, the industry currently lacks a robust solution for multi-model fusion decision-making, primarily due to significant discrepancies between feature extraction and recognition between facial algorithm models from different vendors. Consequently, the industry lacks a robust method for fusion decision-making using algorithms from multiple vendors. Summary of the Invention

[0003] In response to the problems in the prior art, embodiments of the present invention provide a face recognition processing method and apparatus, which can at least partially solve the problems in the prior art.

[0004] In one aspect, the present invention provides a face recognition processing method, comprising:

[0005] Acquire a user's facial image, perform image recognition on the facial image, and obtain scene environment parameters and user characteristic parameters of the facial image;

[0006] Determining the original scores of the models provided by the face recognition multi-model fusion providers according to the scene environment parameters and the user characteristic parameters;

[0007] Each face recognition multi-model fusion provider is rated according to the original score of each model, and the target face recognition multi-model fusion provider that executes the face recognition multi-model fusion decision is determined based on the rating results, and the user's face recognition is performed based on the model provided by the target face recognition multi-model fusion provider.

[0008] The rating of each face recognition multi-model fusion provider based on the original score of each model includes:

[0009] According to the original scores of each model, obtain the standardized matrix and relative entropy probability matrix;

[0010] Based on the relative entropy probability matrix and the normalized matrix, and using the superior-inferior solution distance method evaluation model, the evaluation value of each model corresponding to each face recognition multi-model fusion provider is calculated;

[0011] Each face recognition multi-model fusion provider is rated according to the evaluation value.

[0012] The method of calculating the evaluation value of each model corresponding to each face recognition multi-model fusion provider based on the relative entropy probability matrix and the normalized matrix and using the superior-inferior solution distance method evaluation model includes:

[0013] Calculate a first distance value between each model and the maximum value matrix based on the relative entropy probability matrix, the normalization matrix, and the maximum value matrix of the good-bad solution distance method evaluation model;

[0014] Calculate the second distance value between each model and the minimum value matrix based on the relative entropy probability matrix, the normalization matrix, and the minimum value matrix of the good and bad solution distance method evaluation model;

[0015] Calculate the evaluation scores corresponding to the respective face recognition multi-model fusion providers based on the first distance value and the second distance value;

[0016] Each evaluation score is normalized to obtain the evaluation value corresponding to each model.

[0017] The step of calculating the first distance value between each model and the maximum value matrix based on the relative entropy probability matrix, the normalized matrix, and the maximum value matrix of the good-bad solution distance method evaluation model includes:

[0018] Multiplying the relative entropy probability matrix and the corresponding matrix elements in the normalized matrix respectively to obtain initial data of the good and bad solution distance method evaluation model;

[0019] A first distance value between each model and the maximum value matrix is ​​calculated based on the initial data and the maximum value matrix.

[0020] The step of calculating the second distance value between each model and the minimum value matrix based on the relative entropy probability matrix, the normalized matrix, and the minimum value matrix of the good-bad solution distance method evaluation model includes:

[0021] Multiplying the relative entropy probability matrix and the corresponding matrix elements in the normalized matrix respectively to obtain initial data of the good and bad solution distance method evaluation model;

[0022] A second distance value between each model and the minimum value matrix is ​​calculated based on the initial data and the minimum value matrix.

[0023] The step of obtaining the normalized matrix and the relative entropy probability matrix based on the original scores of each model includes:

[0024] The original scores of each model are used as the elements of the original score matrix to obtain the original score matrix;

[0025] Normalizing the original score matrix to obtain a standardized matrix;

[0026] Relative entropy calculation is performed on the standardized matrix to obtain a relative entropy probability matrix.

[0027] Among them, the target face recognition multi-model fusion provider that performs face recognition multi-model fusion decision is determined based on the rating results, including:

[0028] The face recognition multi-model fusion provider corresponding to the largest evaluation value is determined as the target face recognition multi-model fusion provider.

[0029] In one aspect, the present invention provides a face recognition processing device, comprising:

[0030] an acquisition unit, configured to acquire a facial image of a user, perform image recognition on the facial image, and obtain parameters of a scene environment in which the facial image was captured and user characteristic parameters;

[0031] a determination unit, configured to determine, based on the scene environment parameters and the user characteristic parameters, the original scores of the models provided by the face recognition multi-model fusion providers;

[0032] The recognition unit is used to rate each face recognition multi-model fusion provider according to the original score of each model, determine the target face recognition multi-model fusion provider that executes the face recognition multi-model fusion decision based on the rating results, and perform face recognition on the user based on the model provided by the target face recognition multi-model fusion provider.

[0033] On the other hand, an embodiment of the present invention provides an electronic device, comprising: a processor, a memory, and a bus, wherein:

[0034] The processor and the memory communicate with each other via the bus;

[0035] The memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to execute the following method:

[0036] Acquire a user's facial image, perform image recognition on the facial image, and obtain scene environment parameters and user characteristic parameters of the facial image;

[0037] Determining the original scores of the models provided by the face recognition multi-model fusion providers according to the scene environment parameters and the user characteristic parameters;

[0038] Each face recognition multi-model fusion provider is rated according to the original score of each model, and the target face recognition multi-model fusion provider that executes the face recognition multi-model fusion decision is determined based on the rating results, and the user's face recognition is performed based on the model provided by the target face recognition multi-model fusion provider.

[0039] An embodiment of the present invention provides a non-transitory computer-readable storage medium, including:

[0040] The non-transitory computer-readable storage medium stores computer instructions, which cause the computer to execute the following method:

[0041] Acquire a user's facial image, perform image recognition on the facial image, and obtain scene environment parameters and user characteristic parameters of the facial image;

[0042] Determining the original scores of the models provided by the face recognition multi-model fusion providers according to the scene environment parameters and the user characteristic parameters;

[0043] Each face recognition multi-model fusion provider is rated according to the original score of each model, and the target face recognition multi-model fusion provider that executes the face recognition multi-model fusion decision is determined based on the rating results, and the user's face recognition is performed based on the model provided by the target face recognition multi-model fusion provider.

[0044] The face recognition processing method and device provided by the embodiment of the present invention obtain a user's face image, perform image recognition on the face image, and obtain the scene environment parameters and user characteristic parameters of the shooting of the face image; determine the original score of each model provided by each face recognition multi-model fusion provider according to the scene environment parameters and the user characteristic parameters; rate each face recognition multi-model fusion provider according to the original score of each model, determine the target face recognition multi-model fusion provider that executes the face recognition multi-model fusion decision according to the rating result, and perform face recognition on the user based on the model provided by the target face recognition multi-model fusion provider, which can reasonably select the face recognition multi-model fusion provider that executes the face recognition multi-model fusion decision, thereby ensuring the accuracy of the face recognition result. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0046] Figure 1 The figure is a flowchart of a face recognition processing method provided by one embodiment of the present invention.

[0047] Figure 2 It is a schematic diagram of the modular structure of the face recognition processing method provided by an embodiment of the present invention.

[0048] Figure 3 It is a structural diagram of a face acquisition module provided by an embodiment of the present invention.

[0049] Figure 4 It is a structural diagram of a face data transmission module provided by an embodiment of the present invention.

[0050] Figure 5 It is a structural diagram of the face multi-model fusion decision module provided by an embodiment of the present invention.

[0051] Figure 6 This is a schematic diagram of the data processing flow of the face acquisition module of the face recognition processing method provided by an embodiment of the present invention.

[0052] Figure 7 The present invention provides a face recognition method according to an embodiment of the present invention.

[0053] Figure 8 The present invention provides a face recognition method according to an embodiment of the present invention.

[0054] Figure 9 This is a schematic diagram of the data processing flow of the face multi-model fusion decision module of the face recognition processing method provided by an embodiment of the present invention.

[0055] Figure 10 It is a structural diagram of a face recognition processing device provided by an embodiment of the present invention.

[0056] Figure 11 A schematic diagram of the physical structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0057] To make the purpose, technical solutions and advantages of the embodiments of the present invention more clear, the embodiments of the present invention are further described in detail below with reference to the accompanying drawings. Here, the illustrative embodiments of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention. It should be noted that, unless there is a conflict, the embodiments and features in the embodiments of this application can be combined with each other in any manner.

[0058] Figure 1 FIG. 1 is a flow chart of a face recognition processing method provided by an embodiment of the present invention. Figure 1As shown, the face recognition processing method provided by the embodiment of the present invention includes:

[0059] Step S1: Acquire a user's facial image, perform image recognition on the facial image, and obtain scene environment parameters and user characteristic parameters for shooting the facial image.

[0060] Step S2: Determine the original scores of each model provided by each face recognition multi-model fusion provider based on the scene environment parameters and the user characteristic parameters.

[0061] Step S3: Rating each face recognition multi-model fusion provider according to the original score of each model, determining the target face recognition multi-model fusion provider that executes the face recognition multi-model fusion decision based on the rating results, and performing face recognition on the user based on the model provided by the target face recognition multi-model fusion provider.

[0062] In step S1 above, the device acquires a facial image of a user, performs image recognition on the facial image, and obtains the scene environment parameters and user characteristic parameters of the scene in which the facial image was captured. The device can be a computer device that executes the method, for example, it can include a server. It should be noted that the acquisition and analysis of data involved in the embodiments of the present invention are authorized by the user. When the user conducts a transaction, the user's facial image can be identified based on facial recognition technology to achieve user identity authentication. The facial image can be input into a pre-trained image recognition model, and the scene environment parameters and user characteristic parameters of the facial image can be extracted from the output results of the image recognition model.

[0063] Scene environment parameters may include light intensity and transaction time, etc.

[0064] User characteristic parameters may include whether the user wears a mask, whether the user closes his eyes throughout the process, and whether the user covers his facial features.

[0065] For example: User A uses mobile banking face recognition to log in indoors at 10 am. Set the current scene environment parameter value , user characteristic parameter value .

[0066] Among them, in the scene environment parameters, m1 represents the light intensity, with a value range of 0-indoor scene and 1-outdoor scene; m2 represents the transaction time, with a value range of 0-evening (19:00~07:00), 1-morning (07:00~11:00), 2-noon (11:00~15:00), and 3-afternoon (15:00~19:00).

[0067] Among the user characteristic parameters, n1 indicates whether the user is wearing a mask, with a value range of 0 (not wearing) and 1 (wearing); n2 indicates whether the user has their eyes closed throughout the entire process, with a value range of 0 (not closed) and 1 (closed); and n3 indicates whether the user has covered their facial features, with a value range of 0 (no covering), 1 (covered eyes), 2 (covered nose), and 3 (covered mouth). These scenario environment parameters and user characteristic parameters can be expanded horizontally as the complexity of the scenario increases.

[0068] In step S2, the device determines the raw scores of the models provided by each face recognition multi-model fusion provider based on the scene environment parameters and the user characteristic parameters. Existing model evaluation methods can be used to obtain the raw scores of the models provided by each face recognition multi-model fusion provider by inputting the scene environment parameters and the user characteristic parameters.

[0069] The providers of face recognition multi-model fusion can be simply referred to as Manufacturer A, Manufacturer B, and Manufacturer C. The models provided by each manufacturer can include a liveness detection algorithm, a 1:1 authentication algorithm, and a 1:n recognition algorithm. The 1:1 authentication algorithm and the 1:n recognition algorithm are described as follows:

[0070] 1:1 authentication algorithm: For users with a known unique identity, the algorithm uses the face collected on-site to compare with the face of a user with the same identity that has been registered in the database to confirm whether it is the user.

[0071] 1:n recognition algorithm: For users with unknown unique identities, the algorithm uses the face collected on-site and the n faces of previously registered unique identities in the database for recognition, returns the user's identity and confirms whether it is the user.

[0072] A user with a known unique identity is one who has been verified online by an authoritative organization (identity card number) to confirm that he or she is indeed the user.

[0073] A user with an unknown unique identity is a user whose identity has not been confirmed to be his or her own through online verification by an authoritative organization.

[0074] The raw scores of each model provided by each manufacturer are shown in Table 1:

[0075] Table 1

[0076]

[0077] In the above step S3, the device rates each face recognition multi-model fusion provider according to the original score of each model, determines the target face recognition multi-model fusion provider that executes the face recognition multi-model fusion decision according to the rating results, and performs face recognition on the user based on the model provided by the target face recognition multi-model fusion provider.

[0078] The rating of each face recognition multi-model fusion provider is performed based on the original score of each model, including:

[0079] According to the original scores of each model, a standardized matrix and a relative entropy probability matrix are obtained; the obtaining of the standardized matrix and the relative entropy probability matrix according to the original scores of each model includes:

[0080] The original scores of each model are used as the elements of the original score matrix to obtain the original score matrix;

[0081] Normalizing the original score matrix to obtain a standardized matrix;

[0082] Relative entropy calculation is performed on the standardized matrix to obtain a relative entropy probability matrix.

[0083] According to the relative entropy probability matrix and the standardized matrix, and using the superiority and inferiority solution distance method evaluation model to calculate the evaluation value of each model corresponding to each face recognition multi-model fusion provider; according to the relative entropy probability matrix and the standardized matrix, and using the superiority and inferiority solution distance method evaluation model to calculate the evaluation value of each model corresponding to each face recognition multi-model fusion provider, including:

[0084] The method further comprises: calculating a first distance value between each model and the maximum value matrix according to the relative entropy probability matrix, the standardized matrix, and the maximum value matrix of the good and bad solution distance method evaluation model; and calculating a first distance value between each model and the maximum value matrix according to the relative entropy probability matrix, the standardized matrix, and the maximum value matrix of the good and bad solution distance method evaluation model, including:

[0085] Multiplying the relative entropy probability matrix and the corresponding matrix elements in the normalized matrix respectively to obtain initial data of the good and bad solution distance method evaluation model;

[0086] A first distance value between each model and the maximum value matrix is ​​calculated based on the initial data and the maximum value matrix.

[0087] The method further comprises: calculating a second distance value between each model and the minimum value matrix according to the relative entropy probability matrix, the standardized matrix, and the minimum value matrix of the good and bad solution distance method evaluation model; and calculating a second distance value between each model and the minimum value matrix according to the relative entropy probability matrix, the standardized matrix, and the minimum value matrix of the good and bad solution distance method evaluation model, including:

[0088] Multiplying the relative entropy probability matrix and the corresponding matrix elements in the normalized matrix respectively to obtain initial data of the good and bad solution distance method evaluation model;

[0089] A second distance value between each model and the minimum value matrix is ​​calculated based on the initial data and the minimum value matrix.

[0090] Calculate the evaluation scores corresponding to the respective face recognition multi-model fusion providers based on the first distance value and the second distance value;

[0091] Each evaluation score is normalized to obtain the evaluation value corresponding to each model.

[0092] Each face recognition multi-model fusion provider is rated according to the evaluation value.

[0093] The following are examples:

[0094] make , where X1 represents the parameter decision weight of manufacturer A in the biopsy algorithm, 1:1 authentication algorithm, and 1:n recognition algorithm; X2 represents the parameter decision weight of manufacturer B in the biopsy algorithm, 1:1 authentication algorithm, and 1:n recognition algorithm; and X3 represents the parameter decision weight of manufacturer C in the biopsy algorithm, 1:1 authentication algorithm, and 1:n recognition algorithm. The sub-items x1 to x3 in the parameter weight area represent the processing weight values ​​of the biopsy algorithm, y1 to y3 represent the processing weight values ​​of the face 1:1 authentication algorithm, and z1 to z3 represent the processing weight values ​​of the face 1:n recognition algorithm.

[0095] Using the data in Table 1 as the original score matrix elements, the original score matrix is ​​as follows:

[0096]

[0097] The score range is a positive number from 0 to 100. The original score matrix is ​​normalized and the normalized matrix is ​​denoted as Z. Each element in Z is:

[0098]

[0099] in, Represents the element in the i-th row and j-th column of the original score matrix, and its value range is: , , then the standardized matrix is:

[0100]

[0101] Let the probability matrix of relative entropy be P, then each element in P is:

[0102]

[0103] The probability matrix of relative entropy is:

[0104]

[0105] The initial data of the superior and inferior solution distance method evaluation model .

[0106] The above initial data composition matrix is:

[0107]

[0108] The maximum value matrix of the evaluation model of the good and bad solution distance method is ;

[0109]

[0110] The maximum value matrix of the evaluation model of the good and bad solution distance method is ;

[0111]

[0112] The first distance value between each model and the maximum value matrix is ​​calculated as follows:

[0113]

[0114] The second distance value between each model and the minimum value matrix is ​​calculated as follows:

[0115]

[0116] The evaluation score is calculated as:

[0117]

[0118] The evaluation values ​​of the corresponding models obtained by normalization are:

[0119]

[0120] Referring to the evaluation values ​​of each model in the above example, the normalized evaluation scores are shown in Table 2:

[0121] Table 2

[0122]

[0123] The target face recognition multi-model fusion provider that performs face recognition multi-model fusion decision-making is determined based on the rating results, including:

[0124] The face recognition multi-model fusion provider corresponding to the largest numerical evaluation value is determined as the target face recognition multi-model fusion provider. Referring to Table 2, manufacturer A is determined as the target face recognition multi-model fusion provider. Performing face recognition on the user based on the model provided by the target face recognition multi-model fusion provider may include:

[0125] First, determine whether face recognition is performed through the liveness detection algorithm. If so, face recognition is performed directly through the liveness detection algorithm, and the method terminates execution.

[0126] If not, it is determined whether face recognition is performed through the 1:1 authentication algorithm. If yes, face recognition is performed directly through the 1:1 authentication algorithm, and the method terminates execution.

[0127] If not, it is determined whether face recognition is performed using the 1:n recognition algorithm. If so, face recognition is performed using the 1:1 authentication algorithm, and the method terminates execution.

[0128] like Figure 2 As shown, the face recognition processing method provided by the embodiment of the present invention can be implemented based on modularization, specifically including: a face recognition main control module 1, a face acquisition module 2, a face data transmission module 3 and a face multi-model fusion decision module 4; wherein:

[0129] When conducting a face recognition transaction, the face acquisition module 2 drives the camera to collect the user's face image / video data, and collects the current scene environment parameters and user characteristic parameters. The face image / video data is uploaded to the face multi-model fusion decision module 4 through the face data transmission module 3 to complete the parameter template selection of the transaction in this scenario.

[0130] like Figure 3 As shown, the face acquisition module 2 includes a face acquisition main control MCU unit 21, a face data acquisition control unit 22, a face data quality control unit 23 and a scene parameter and user characteristic parameter acquisition unit 24; wherein:

[0131] The face acquisition module 2 primarily performs facial image / video data acquisition, facial quality detection and control, and environmental parameter and user characteristic parameter collection. The face acquisition master MCU unit 21 drives the camera deployed on the intelligent interactive device to collect the user's facial image / video data, and drives the environmental sensor to collect data such as ambient light intensity, background depth, face distance, multiple face situations, pitch angle, and left and right tilt angle. The face acquisition master MCU unit 21 sends the collected facial image / video data to the facial data acquisition control unit 22, facial data quality control unit 23, and scene parameter and user characteristic parameter acquisition unit 24. When the collected data is an image, it directly determines whether the face in the image meets the usage standards: interpupillary distance >= 60 pixels, whether there is motion blur, whether the image is too bright, too dark, or no face. When the collected data is a video, it first extracts key frames from the video data. The video duration must be less than 5 seconds. To improve efficiency, 2 frames / s are extracted as key frames for detection. It then determines whether the face in the image frame meets the usage standards: interpupillary distance >= 60 pixels, whether there is motion blur, whether the image is too bright, too dark, or no face. If the requirements are met, it initiates a request to transmit and process the facial data and environmental parameters.

[0132] like Figure 4 As shown, the face data transmission module 3 includes a face data transmission main control unit 31, a data upload unit 32 and a data sending unit 33; wherein:

[0133] The facial data transmission module 3 primarily handles uploading facial images / video data and environmental parameters, and distributing facial images / features. The facial data transmission control unit 31 first performs a data security check to determine whether the uploaded data is facial images / video data and environmental parameters. If so, the facial data is uploaded to the cloud database via the data upload unit 32. When downloading data, the data distributing unit 33 directly initiates an image / feature query request to the cloud database. Once the query results are obtained, the data distributing unit 33 distributes them to the intelligent interactive device or application server.

[0134] like Figure 5 As shown, the face multi-model fusion decision module 4 includes a face recognition algorithm main control unit 41, a face multi-model fusion decision unit 42, and the face multi-model fusion decision unit 42 includes: a face multi-model scheduling unit 421, a face multi-model fusion unit 422, a face model A423, a face model B424, and a face model C425, where A, B, and C respectively refer to different versions of face algorithm models from different manufacturers, which can be expanded horizontally and vertically and are not limited to the three examples in this block diagram. The face multi-model fusion decision module 4 also includes a face registration algorithm processing unit 43, a face 1:1 algorithm processing unit 44, and a face 1:n algorithm processing unit 45; wherein:

[0135] The multi-model face fusion decision module 4 primarily selects model fusion parameters for specific scenarios and transactions, registers faces based on the parameters and weights of a single model, and processes the face recognition algorithm. Ultimately, it returns the registration / recognition results to the intelligent interactive device front end. First, it schedules the face models, retrieves the type and number of algorithm models on the server side, and sends the scheduling results to the multi-model face fusion decision module 4. The module then uses the scenario parameters and user characteristic parameters to fit a fusion parameter template for that scenario.

[0136] Multi-model fusion decision-making is mainly achieved through Stacking ensemble learning, which combines different face recognition models as base learners through algorithms such as XGBoost and LSTM to obtain better recognition performance than a single algorithm. Stacking ensemble learning not only analyzes the individual recognition capabilities of each base learner, but also comprehensively compares the combined effects of each base learner, and fits the optimal parameter configuration template for the scenario through scenario parameters and user characteristic parameters.

[0137] Stacking ensemble learning uses a deep learning network (DNN) for adaptive training. Sample data is comprehensively derived from data such as ambient light intensity, background depth, face distance, multiple faces, pitch and tilt angles, and left and right tilt angles in various face scenarios to achieve parameter fitting.

[0138] After obtaining the optimal parameter template for model fusion, enter the face algorithm processing module of a single model, first determine whether the processed data is an image, if it is an image, directly send it to the corresponding algorithm processing unit, if it is a video, first extract the best face frame, and then send it to the corresponding algorithm processing unit.

[0139] The facial image data is fed into the corresponding algorithm service processing unit for processing: For data requiring face registration algorithm processing, data preprocessing is performed, facial feature extraction and modeling are completed, and the user information and facial features are then registered with the face recognition system database. For requests requiring 1:1 face recognition algorithm processing, image preprocessing is performed, facial feature extraction and modeling are completed, and then the feature data is compared 1:1 with the facial features retrieved in the database, and the comparison results are returned to the requesting client.

[0140] For those that need to be processed by 1:n face recognition algorithm, after completing image preprocessing, facial feature extraction and modeling, identity recognition is performed with n faces of known user unique identities in the database, and the recognition results are returned to the requesting client.

[0141] like Figure 6 As shown, the data processing process of the face acquisition module of the face recognition processing method provided by the embodiment of the present invention is described as follows:

[0142] S101: Initiate a face recognition data collection request;

[0143] S102: Environmental sensors collect scene environmental parameters;

[0144] S103: driving the camera to collect facial image / video data;

[0145] S104: face data quality judgement;

[0146] S105: Is the data collected face image data?

[0147] S106: Collecting facial video data?

[0148] S107: Extract video key frames and detect them frame by frame

[0149] S108: Is the interpupillary distance between the eyes not meeting the requirements? Or is there motion blur? Or is the image too bright? Or is the image too dark? Or is there no face?

[0150] S109: User characteristic parameter analyzer;

[0151] S110: Initiate a request to upload facial data;

[0152] S111: The face data collection process ends.

[0153] like Figure 7 As shown, the upload data processing process of the face data transmission module of the face recognition processing method provided by the embodiment of the present invention is described as follows:

[0154] S201: Initiate a facial data upload request;

[0155] S202: Data security control unit;

[0156] S203: Is the uploaded data facial image data or video data?

[0157] S204: facial data upload request rejected;

[0158] S205: The face data upload process ends;

[0159] S206: Do you want to upload face registration image data or video data?

[0160] S207: Data is uploaded to the cloud database via https / Socket;

[0161] S208: Initiate facial algorithm intelligent decision-making and registration algorithm processing;

[0162] S209: Face registration data upload completed;

[0163] S210: The face data upload process ends;

[0164] S211: Do you want to upload 1:1 facial recognition image data or video data?

[0165] S212: Data is uploaded to the cloud database via https / Socket;

[0166] S213: Initiate intelligent decision-making of facial algorithms and 1:1 recognition algorithm processing;

[0167] S214: 1:1 face data upload completed;

[0168] S215: The face data upload process ends;

[0169] S216: Do you want to upload face 1:n recognition image data or video data?

[0170] S217: Data is uploaded to the cloud database via https / Socket;

[0171] S218: Initiate intelligent decision-making of face algorithm and 1:n recognition algorithm processing;

[0172] S219: Face 1:n data upload completed;

[0173] S220: The face data upload process ends;

[0174] S221: facial data upload request rejected;

[0175] S222: The facial data upload process ends.

[0176] like Figure 8 As shown, the downlink data processing process of the face data transmission module of the face recognition processing method provided by the embodiment of the present invention is described as follows:

[0177] S301: Face data sending request;

[0178] S302: Is the data sent facial feature data or facial image data?

[0179] S303: Request for facial data delivery is rejected;

[0180] S304: The facial data delivery process ends;

[0181] S305: Should 1:1 facial recognition feature data or image data be sent?

[0182] S306: Obtaining facial feature data or image data from a cloud database based on the customer information;

[0183] S307: Sending a facial feature or image to a designated interactive device / application server;

[0184] S308: Face data sending request completed;

[0185] S309: The facial data delivery process ends;

[0186] S310: Should facial 1:n recognition feature data or image data be sent?

[0187] S311: Obtain N facial feature data or image data from the cloud database based on the customer group information;

[0188] S312: Sending N facial features or images to a designated interactive device / application server;

[0189] S313: Face data sending request completed;

[0190] S314: The facial data delivery process ends;

[0191] S315: Request for facial data delivery rejected;

[0192] S316: The facial data sending process ends.

[0193] like Figure 9 As shown, the data processing process of the face multi-model fusion decision module of the face recognition processing method provided by the embodiment of the present invention is described as follows:

[0194] S401: Initiate a face recognition algorithm processing request;

[0195] S402: Perform multi-model face scheduling;

[0196] S403: calling face multi-model fusion decision;

[0197] S404: Fitting an algorithm parameter template based on environmental parameters and user characteristics;

[0198] S405: Send the fitted parameter template to the corresponding algorithm module for processing;

[0199] S406: Determine whether the processed data is an image.

[0200] S407: Determine whether the processed data is a video.

[0201] S408: Face algorithm processing request rejected;

[0202] S409: The face algorithm processing process ends;

[0203] S410: Extracting the best video frame from the video data;

[0204] S411: Do you want to perform face registration algorithm processing request?

[0205] S412: face registration algorithm processing unit;

[0206] S413: Face image / video data preprocessing;

[0207] S414: performing facial feature extraction and modeling for different algorithm models according to the multi-model parameter template;

[0208] S415: Saving facial features to a cloud database;

[0209] S416: Return the registration result to the requesting client;

[0210] S417: Face registration algorithm processing request completed;

[0211] S418: The face algorithm processing process ends;

[0212] S419: Do you want to perform a 1:1 face recognition algorithm processing request?

[0213] S420: face 1:1 recognition algorithm processing unit;

[0214] S421: Face image / video data preprocessing;

[0215] S422: extracting and modeling facial features based on multi-model parameter templates for different algorithm models;

[0216] S423: Perform a 1:1 comparison between different algorithm models and a facial feature acquired from the cloud based on the multi-model parameter template;

[0217] S424: Calculate the comparison result based on the weight and return it to the requesting client;

[0218] S425: 1:1 face recognition algorithm processing request completed;

[0219] S426: The face algorithm processing process ends;

[0220] S427: Do you want to perform face 1:n recognition algorithm processing request?

[0221] S428: face 1:n recognition algorithm processing unit;

[0222] S429: Face image / video data preprocessing;

[0223] S430: performing facial feature extraction and modeling for different algorithm models according to the multi-model parameter template;

[0224] S431: Perform 1:n recognition on N facial features obtained from the cloud using different algorithm models according to the multi-model parameter template;

[0225] S432: Calculate the comparison result based on the weight and return the recognition result to the requesting client;

[0226] S433: Face 1:n recognition algorithm processing request completed;

[0227] S434: The face algorithm processing process ends;

[0228] S435: Face algorithm processing request rejected;

[0229] S436 The face algorithm processing process ends.

[0230] The face recognition processing method provided by the embodiment of the present invention obtains a face image of a user, performs image recognition on the face image, and obtains scene environment parameters and user characteristic parameters for shooting the face image; determines the original scores of each model provided by each face recognition multi-model fusion provider according to the scene environment parameters and the user characteristic parameters; rates each face recognition multi-model fusion provider according to the original scores of each model, determines the target face recognition multi-model fusion provider that executes the face recognition multi-model fusion decision according to the rating results, and performs face recognition on the user based on the model provided by the target face recognition multi-model fusion provider. The face recognition multi-model fusion provider that executes the face recognition multi-model fusion decision can be reasonably selected, thereby ensuring the accuracy of the face recognition result.

[0231] Furthermore, the rating of each face recognition multi-model fusion provider based on the original score of each model includes:

[0232] According to the original scores of each model, the standardized matrix and relative entropy probability matrix are obtained; please refer to the above description and will not be repeated here.

[0233] According to the relative entropy probability matrix and the standardized matrix, the evaluation model is calculated using the superior and inferior solution distance method and the evaluation values ​​of each model corresponding to each face recognition multi-model fusion provider are respectively calculated; please refer to the above description and no further details will be given.

[0234] Each face recognition multi-model fusion provider is rated based on the evaluation value. Please refer to the above description and do not repeat it here.

[0235] The face recognition processing method provided by the embodiment of the present invention uses the superior and inferior solution distance method to evaluate the model and calculate the evaluation value, and can further reasonably select the face recognition multi-model fusion provider that executes the face recognition multi-model fusion decision, thereby ensuring the accuracy of the face recognition result.

[0236] Furthermore, the calculation of the evaluation value of each model corresponding to each face recognition multi-model fusion provider based on the relative entropy probability matrix and the normalized matrix and using the superior-inferior solution distance method evaluation model includes:

[0237] According to the relative entropy probability matrix, the standardized matrix and the maximum value matrix of the superior and inferior solution distance method evaluation model, the first distance value between each model and the maximum value matrix is ​​calculated; please refer to the above description and no further details will be given.

[0238] According to the relative entropy probability matrix, the standardized matrix and the minimum value matrix of the good and bad solution distance method evaluation model, the second distance value between each model and the minimum value matrix is ​​calculated; please refer to the above description and no further details will be given.

[0239] According to the first distance value and the second distance value, the evaluation scores corresponding to each face recognition multi-model fusion provider are calculated; please refer to the above description and no further details will be given.

[0240] Each evaluation score is normalized to obtain the evaluation value of each model. Please refer to the above description and do not repeat it here.

[0241] The face recognition processing method provided by the embodiment of the present invention can further reasonably select the face recognition multi-model fusion provider that executes the face recognition multi-model fusion decision by calculating the first distance value and the second distance value, thereby ensuring the accuracy of the face recognition result.

[0242] Furthermore, the step of calculating the first distance value between each model and the maximum value matrix based on the relative entropy probability matrix, the normalized matrix, and the maximum value matrix of the evaluation model using the superior and inferior solution distance method includes:

[0243] The relative entropy probability matrix and the corresponding matrix elements in the standardized matrix are multiplied respectively to obtain the initial data of the good and bad solution distance method evaluation model; please refer to the above description and no further details will be given.

[0244] Based on the initial data and the maximum value matrix, a first distance value between each model and the maximum value matrix is ​​calculated.

[0245] The face recognition processing method provided by the embodiment of the present invention can quickly calculate the first distance value.

[0246] Furthermore, the step of calculating the second distance value between each model and the minimum value matrix based on the relative entropy probability matrix, the normalized matrix, and the minimum value matrix of the evaluation model using the superior and inferior solution distance method includes:

[0247] The relative entropy probability matrix and the corresponding matrix elements in the standardized matrix are multiplied respectively to obtain the initial data of the good and bad solution distance method evaluation model; please refer to the above description and no further details will be given.

[0248] According to the initial data and the minimum value matrix, a second distance value between each model and the minimum value matrix is ​​calculated.

[0249] The face recognition processing method provided by the embodiment of the present invention can quickly calculate the second distance value.

[0250] Furthermore, the step of obtaining a normalized matrix and a relative entropy probability matrix based on the original scores of each model includes:

[0251] The original scores of each model are used as elements of the original score matrix to obtain the original score matrix; please refer to the above description and will not repeat it here.

[0252] The original score matrix is ​​normalized to obtain a normalized matrix; please refer to the above description and will not be repeated here.

[0253] The relative entropy is calculated on the standardized matrix to obtain a relative entropy probability matrix.

[0254] The face recognition processing method provided by the embodiment of the present invention can quickly calculate the normalization matrix and the relative entropy probability matrix.

[0255] Furthermore, the target face recognition multi-model fusion provider for executing the face recognition multi-model fusion decision is determined based on the rating results, including:

[0256] The face recognition multi-model fusion provider corresponding to the largest evaluation value is determined as the target face recognition multi-model fusion provider. Please refer to the above description and do not repeat it here.

[0257] The face recognition processing method provided by the embodiment of the present invention can further reasonably select a face recognition multi-model fusion provider that executes the face recognition multi-model fusion decision, thereby ensuring the accuracy of the face recognition result.

[0258] It should be noted that the face recognition processing method provided in the embodiment of the present invention can be used in the financial field, and can also be used in any technical field other than the financial field. The embodiment of the present invention does not limit the application field of the face recognition processing method.

[0259] Figure 10 FIG. 1 is a structural diagram of a face recognition processing device provided by an embodiment of the present invention. Figure 10As shown, the face recognition processing device provided by the embodiment of the present invention includes an acquisition unit 1001, a determination unit 1002 and a recognition unit 1003, wherein:

[0260] The acquisition unit 1001 is used to acquire the user's facial image, perform image recognition on the facial image, and obtain the scene environment parameters and user characteristic parameters for shooting the facial image; the determination unit 1002 is used to determine the original scores of each model provided by each face recognition multi-model fusion provider based on the scene environment parameters and the user characteristic parameters; the identification unit 1003 is used to rate each face recognition multi-model fusion provider based on the original scores of each model, determine the target face recognition multi-model fusion provider that executes the face recognition multi-model fusion decision based on the rating results, and perform face recognition on the user based on the model provided by the target face recognition multi-model fusion provider.

[0261] Specifically, the acquisition unit 1001 in the device is used to obtain the user's facial image, perform image recognition on the facial image, and obtain the scene environment parameters and user characteristic parameters for shooting the facial image; the determination unit 1002 is used to determine the original scores of each model provided by each face recognition multi-model fusion provider based on the scene environment parameters and the user characteristic parameters; the identification unit 1003 is used to rate each face recognition multi-model fusion provider based on the original scores of each model, determine the target face recognition multi-model fusion provider that executes the face recognition multi-model fusion decision based on the rating results, and perform face recognition on the user based on the model provided by the target face recognition multi-model fusion provider.

[0262] The face recognition processing device provided by the embodiment of the present invention obtains a face image of a user, performs image recognition on the face image, and obtains scene environment parameters and user characteristic parameters for shooting the face image; determines the original scores of each model provided by each face recognition multi-model fusion provider according to the scene environment parameters and the user characteristic parameters; rates each face recognition multi-model fusion provider according to the original scores of each model, determines the target face recognition multi-model fusion provider that executes the face recognition multi-model fusion decision according to the rating results, and performs face recognition on the user based on the model provided by the target face recognition multi-model fusion provider. The face recognition multi-model fusion provider that executes the face recognition multi-model fusion decision can be reasonably selected, thereby ensuring the accuracy of the face recognition result.

[0263] Furthermore, the identification unit 1003 is specifically configured to:

[0264] According to the original scores of each model, obtain the standardized matrix and relative entropy probability matrix;

[0265] Based on the relative entropy probability matrix and the normalized matrix, and using the superior-inferior solution distance method evaluation model, the evaluation value of each model corresponding to each face recognition multi-model fusion provider is calculated;

[0266] Each face recognition multi-model fusion provider is rated according to the evaluation value.

[0267] The face recognition processing device provided by the embodiment of the present invention uses the superior and inferior solution distance method evaluation model to calculate the evaluation value, and can further reasonably select the face recognition multi-model fusion provider that executes the face recognition multi-model fusion decision, thereby ensuring the accuracy of the face recognition result.

[0268] Furthermore, the identification unit 1003 is further specifically configured to:

[0269] Calculate a first distance value between each model and the maximum value matrix based on the relative entropy probability matrix, the normalization matrix, and the maximum value matrix of the good-bad solution distance method evaluation model;

[0270] Calculate the second distance value between each model and the minimum value matrix based on the relative entropy probability matrix, the normalization matrix, and the minimum value matrix of the good and bad solution distance method evaluation model;

[0271] Calculate the evaluation scores corresponding to the respective face recognition multi-model fusion providers based on the first distance value and the second distance value;

[0272] Each evaluation score is normalized to obtain the evaluation value corresponding to each model.

[0273] The face recognition processing device provided by the embodiment of the present invention can further reasonably select the face recognition multi-model fusion provider that executes the face recognition multi-model fusion decision by calculating the first distance value and the second distance value, thereby ensuring the accuracy of the face recognition result.

[0274] Furthermore, the identification unit 1003 is further specifically configured to:

[0275] Multiplying the relative entropy probability matrix and the corresponding matrix elements in the normalized matrix respectively to obtain initial data of the good and bad solution distance method evaluation model;

[0276] A first distance value between each model and the maximum value matrix is ​​calculated based on the initial data and the maximum value matrix.

[0277] The face recognition processing device provided by the embodiment of the present invention can quickly calculate the first distance value.

[0278] Furthermore, the identification unit 1003 is further specifically configured to:

[0279] Multiplying the relative entropy probability matrix and the corresponding matrix elements in the normalized matrix respectively to obtain initial data of the good and bad solution distance method evaluation model;

[0280] A second distance value between each model and the minimum value matrix is ​​calculated based on the initial data and the minimum value matrix.

[0281] The face recognition processing device provided by the embodiment of the present invention can quickly calculate the second distance value.

[0282] Furthermore, the identification unit 1003 is further specifically configured to:

[0283] The original scores of each model are used as the elements of the original score matrix to obtain the original score matrix;

[0284] Normalizing the original score matrix to obtain a standardized matrix;

[0285] Relative entropy calculation is performed on the standardized matrix to obtain a relative entropy probability matrix.

[0286] The face recognition processing device provided by the embodiment of the present invention can quickly calculate the normalization matrix and the relative entropy probability matrix.

[0287] Furthermore, the identification unit 1003 is specifically configured to:

[0288] The face recognition multi-model fusion provider corresponding to the largest evaluation value is determined as the target face recognition multi-model fusion provider.

[0289] The face recognition processing device provided by the embodiment of the present invention can further reasonably select a face recognition multi-model fusion provider that executes the face recognition multi-model fusion decision, thereby ensuring the accuracy of the face recognition result.

[0290] The embodiment of the face recognition processing device provided in the embodiment of the present invention can be specifically used to execute the processing flow of the above-mentioned method embodiments. Its functions are not described in detail here, and reference can be made to the detailed description of the above-mentioned method embodiments.

[0291] Figure 11 A schematic diagram of the physical structure of an electronic device provided in an embodiment of the present invention, such as Figure 11 As shown, the electronic device includes: a processor (processor) 1101, a memory (memory) 1102 and a bus 1103;

[0292] The processor 1101 and the memory 1102 communicate with each other via a bus 1103.

[0293] The processor 1101 is configured to call the program instructions in the memory 1102 to execute the methods provided by the above method embodiments, for example, including:

[0294] Acquire a user's facial image, perform image recognition on the facial image, and obtain scene environment parameters and user characteristic parameters of the facial image;

[0295] Determining the original scores of the models provided by the face recognition multi-model fusion providers according to the scene environment parameters and the user characteristic parameters;

[0296] Each face recognition multi-model fusion provider is rated according to the original score of each model, and the target face recognition multi-model fusion provider that executes the face recognition multi-model fusion decision is determined based on the rating results, and the user's face recognition is performed based on the model provided by the target face recognition multi-model fusion provider.

[0297] This embodiment discloses a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium. The computer program includes program instructions. When the program instructions are executed by a computer, the computer can perform the methods provided in the above-mentioned method embodiments, for example, including:

[0298] Acquire a user's facial image, perform image recognition on the facial image, and obtain scene environment parameters and user characteristic parameters of the facial image;

[0299] Determining the original scores of the models provided by the face recognition multi-model fusion providers according to the scene environment parameters and the user characteristic parameters;

[0300] Each face recognition multi-model fusion provider is rated according to the original score of each model, and the target face recognition multi-model fusion provider that executes the face recognition multi-model fusion decision is determined based on the rating results, and the user's face recognition is performed based on the model provided by the target face recognition multi-model fusion provider.

[0301] This embodiment provides a computer-readable storage medium storing a computer program. The computer program enables the computer to execute the methods provided in the above method embodiments, for example, including:

[0302] Acquire a user's facial image, perform image recognition on the facial image, and obtain scene environment parameters and user characteristic parameters of the facial image;

[0303] Determining the original scores of the models provided by the face recognition multi-model fusion providers according to the scene environment parameters and the user characteristic parameters;

[0304] Each face recognition multi-model fusion provider is rated according to the original score of each model, and the target face recognition multi-model fusion provider that executes the face recognition multi-model fusion decision is determined based on the rating results, and the user's face recognition is performed based on the model provided by the target face recognition multi-model fusion provider.

[0305] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, 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 magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0306] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0307] These computer program instructions may 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 produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0308] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0309] Throughout this specification, reference to terms such as "one embodiment," "a specific embodiment," "some embodiments," "for example," "example," "specific example," or "some examples" means that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0310] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A face recognition processing method, characterized in that: include: Acquire a facial image of the user, perform image recognition on the facial image, and obtain scene environment parameters and user characteristic parameters of the scene in which the facial image was captured, wherein the scene environment parameters include light intensity and transaction time, and the user characteristic parameters include whether the user wears a mask, whether the user closes their eyes throughout the transaction, and whether the user covers their facial features; Determining the original scores of the models provided by the face recognition multi-model fusion providers according to the scene environment parameters and the user characteristic parameters; Rating each face recognition multi-model fusion provider according to the original score of each model, determining the target face recognition multi-model fusion provider that executes the face recognition multi-model fusion decision based on the rating results, and performing face recognition on the user based on the model provided by the target face recognition multi-model fusion provider; The rating of each face recognition multi-model fusion provider based on the original score of each model includes: The original scores of each model are used as the elements of the original score matrix to obtain the original score matrix; Normalizing the original score matrix to obtain a standardized matrix; Performing relative entropy calculation on the standardized matrix to obtain a relative entropy probability matrix; Based on the relative entropy probability matrix and the normalized matrix, and using the superior-inferior solution distance method evaluation model, the evaluation value of each model corresponding to each face recognition multi-model fusion provider is calculated; Each face recognition multi-model fusion provider is rated according to the evaluation value.

2. The face recognition processing method according to claim 1, characterized in that: The method of calculating the evaluation value of each model corresponding to each face recognition multi-model fusion provider based on the relative entropy probability matrix and the normalized matrix and using the superior-inferior solution distance method evaluation model includes: Calculate a first distance value between each model and the maximum value matrix based on the relative entropy probability matrix, the normalization matrix, and the maximum value matrix of the good-bad solution distance method evaluation model; Calculate the second distance value between each model and the minimum value matrix based on the relative entropy probability matrix, the normalization matrix, and the minimum value matrix of the good and bad solution distance method evaluation model; Calculate the evaluation scores corresponding to the respective face recognition multi-model fusion providers based on the first distance value and the second distance value; Each evaluation score is normalized to obtain the evaluation value corresponding to each model.

3. The face recognition processing method according to claim 2, characterized in that: The step of calculating a first distance value between each model and the maximum value matrix based on the relative entropy probability matrix, the normalized matrix, and the maximum value matrix of the superior and inferior solution distance method evaluation model includes: Multiplying the relative entropy probability matrix and the corresponding matrix elements in the normalized matrix respectively to obtain initial data of the good and bad solution distance method evaluation model; A first distance value between each model and the maximum value matrix is ​​calculated based on the initial data and the maximum value matrix.

4. The face recognition processing method according to claim 2, characterized in that: The step of calculating a second distance value between each model and the minimum value matrix based on the relative entropy probability matrix, the normalized matrix, and the minimum value matrix of the evaluation model using the superior and inferior solution distance method comprises: Multiplying the relative entropy probability matrix and the corresponding matrix elements in the normalized matrix respectively to obtain initial data of the good and bad solution distance method evaluation model; A second distance value between each model and the minimum value matrix is ​​calculated based on the initial data and the minimum value matrix.

5. The face recognition processing method according to any one of claims 1 to 4, characterized in that: The target face recognition multi-model fusion provider that performs face recognition multi-model fusion decision-making is determined based on the rating results, including: The face recognition multi-model fusion provider corresponding to the largest evaluation value is determined as the target face recognition multi-model fusion provider.

6. A face recognition processing device, characterized in that: include: an acquisition unit, configured to acquire a facial image of the user, perform image recognition on the facial image, and obtain scene environment parameters and user characteristic parameters of the scene in which the facial image was captured, wherein the scene environment parameters include light intensity and transaction time, and the user characteristic parameters include whether the user wears a mask, whether the user closes his eyes throughout the transaction, and whether the user covers his facial features; a determination unit, configured to determine, based on the scene environment parameters and the user characteristic parameters, the original scores of the models provided by the face recognition multi-model fusion providers; an identification unit, configured to rate each face recognition multi-model fusion provider based on the original score of each model, determine a target face recognition multi-model fusion provider for executing the face recognition multi-model fusion decision based on the rating results, and perform face recognition on the user based on the model provided by the target face recognition multi-model fusion provider; Among them, the recognition unit is specifically used to use the original score of each model as an element of the original score matrix to obtain an original score matrix; standardize the original score matrix to obtain a standardized matrix; perform relative entropy calculation on the standardized matrix to obtain a relative entropy probability matrix; based on the relative entropy probability matrix and the standardized matrix, and using the superiority and inferior solution distance method to evaluate the model calculation and the evaluation value corresponding to each model of each face recognition multi-model fusion provider; and rate each face recognition multi-model fusion provider according to each evaluation value.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 5 are implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Images