Identity authentication method and device

Through the combination of deep convolutional neural network and GA-BP neural network, the facial image features are extracted and the cosine similarity is calculated, which solves the accuracy problem of the existing identity authentication method when faces differ greatly or expire during the validity period of the ID card. , improves the efficiency of identity authentication and banking business processing efficiency.

CN113642468BInactive Publication Date: 2024-11-26BANK OF CHINA
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202110936018.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-08-16
Publication Date
2024-11-26
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The existing identity authentication method compares whether the face image on the user's ID card is consistent with the user's face image collected on the spot, resulting in inaccurate verification in the case where the face image is large within the validity period of the ID card, and Identity authentication cannot be performed when the ID card expires, which affects banking business efficiency and user experience.

Method used

A pre-trained deep convolutional neural network is used to extract the feature vectors of the current face image and the face image on the ID card, and a GA-BP neural network is used to determine whether the banking business can be handled based on the cosine similarity and user identity information, providing more accurate the identity authentication result.

Benefits of technology

It improves the accuracy and efficiency of identity authentication, allows the user's identity to be verified when the ID card expires, and avoids identity authentication failure when the facial images are large, improving banking business processing efficiency and user experience.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN113642468B_ABST
    Figure CN113642468B_ABST
Patent Text Reader

Abstract

The present invention discloses an identity authentication method and device, which relates to the field of artificial intelligence technology; wherein the method comprises: collecting the current face image of the user and the face image on the user's ID card; using a pre-trained deep convolutional neural network to extract the first feature vector of the current face image and the second feature vector of the face image on the user's ID card; calculating the cosine similarity between the current face image of the user and the face image on the user's ID card according to the first feature vector and the second feature vector; inputting the bank service to be handled by the user, the cosine similarity, and the identity information of the user into the pre-trained GA-BP neural network to obtain a verification result of whether the user can handle the bank service. The present invention improves the accuracy and efficiency of identity authentication, thereby improving the handling efficiency of bank services and improving the user experience.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an identity authentication method and device. Background Art

[0002] This section is intended to provide a background or context to the embodiments of the invention recited in the claims. No admission is made that the description herein is prior art by inclusion in this section.

[0003] At present, when a user goes to a bank to handle banking business, he needs to use his ID card to authenticate the user. Only when the authentication is passed can the corresponding banking business be handled for the user. When the ID card is used to authenticate the user, the facial image on the user's ID card is compared with the user's facial image collected on site to see if they are consistent. If the comparison result is consistent, the user passes the authentication. If the comparison result is inconsistent, the user fails the authentication. Because this verification method compares the facial image on the user's ID card with the user's facial image collected on site, in order to avoid the facial image on the user's ID card and the facial image collected on site being very different, the ID card can only be authenticated within the validity period. However, if the ID card is within the validity period but has not been replaced for a long time, the user's facial image and the facial image on the ID card may also be very different. Therefore, the current authentication method only authenticates the user by comparing the facial image on the user's ID card with the facial image collected on site to see if they are consistent, resulting in inaccurate authentication results. In addition, when the ID card expires, the user cannot authenticate the user, resulting in low authentication efficiency, which affects the efficiency of banking business and poor user experience. Summary of the invention

[0004] The embodiment of the present invention provides an identity authentication method for improving the efficiency of identity authentication, thereby improving the efficiency of handling banking services and improving user experience. The method includes:

[0005] Collect the user's current face image and the face image on the user's ID card;

[0006] Extracting a first feature vector of the current face image and a second feature vector of the face image on the user's ID card using a pre-trained deep convolutional neural network; wherein the deep convolutional neural network is used to filter the input image using at least one convolutional layer to obtain a feature vector of the image;

[0007] Calculating the cosine similarity between the current face image of the user and the face image on the user's ID card based on the first feature vector and the second feature vector;

[0008] The banking service that the user wants to handle, the cosine similarity, and the user's identity information are input into a pre-trained GA-BP neural network to obtain a verification result of whether the user can handle the banking service; wherein the GA-BP neural network is used to authenticate the user's identity and determine whether the user can handle the input banking service.

[0009] The embodiment of the present invention further provides an identity authentication device for improving the efficiency of identity authentication, thereby improving the efficiency of handling banking services and improving user experience. The device includes:

[0010] The acquisition module is used to acquire the current face image of the user and the face image on the user's ID card;

[0011] An extraction module, used to extract a first feature vector of the current face image and a second feature vector of the face image on the user's ID card using a pre-trained deep convolutional neural network; wherein the deep convolutional neural network is used to filter the input image using at least one convolutional layer to obtain a feature vector of the image;

[0012] A calculation module, used for calculating the cosine similarity between the current face image of the user and the face image on the user's ID card according to the first feature vector and the second feature vector;

[0013] The processing module is used to input the banking business to be handled by the user, the cosine similarity, and the user's identity information into a pre-trained GA-BP neural network to obtain a verification result of whether the user can handle the banking business; wherein the GA-BP neural network is used to authenticate the user's identity and determine whether the user can handle the input banking business.

[0014] An embodiment of the present invention further provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned identity authentication method when executing the computer program.

[0015] An embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program for executing the above-mentioned identity authentication method.

[0016] In an embodiment of the present invention, a current facial image of a user and a facial image on the user's ID card are collected; a first feature vector of the current facial image and a second feature vector of the facial image on the user's ID card are extracted using a pre-trained deep convolutional neural network; wherein the deep convolutional neural network is used to filter an input image using at least one convolutional layer to obtain a feature vector of the image; the cosine similarity between the current facial image of the user and the facial image on the user's ID card is calculated based on the first feature vector and the second feature vector; the banking business to be handled by the user, the cosine similarity, and the user's identity information are input into the pre-trained GA-BP neural network to obtain a verification result of whether the user can handle the banking business; wherein the GA-BP neural network is used to authenticate the user's identity and determine whether the user can handle the input banking business. In this way, the GA-BP neural network can be used to obtain the verification result of whether the user can handle the bank business according to the user's identity information, cosine similarity, and the bank business to be handled. It is not only judged by the similarity between the face image on the user's ID card and the user's current face image. Therefore, the user can also verify whether he can handle the bank business when the ID card is expired. And when there is a large difference between the face image on the ID card and the user's current face image, it will not directly lead to the failure of the user's identity authentication and the inability to handle the bank business to be handled. The accuracy and efficiency of identity authentication are improved, thereby improving the efficiency of bank business handling and improving user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] 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 use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0018] Figure 1 A flowchart of an identity authentication method provided in an embodiment of the present invention;

[0019] Figure 2 A schematic diagram of the structure of a deep convolutional neural network provided in an embodiment of the present invention;

[0020] Figure 3 A schematic diagram of the structure of another deep convolutional neural network provided in an embodiment of the present invention;

[0021] Figure 4 This is a structural example diagram of a GA-BP neural network provided in an embodiment of the present invention;

[0022] Figure 5A flow chart of a method for training a GA-BP neural network provided in an embodiment of the present invention;

[0023] Figure 6 A schematic diagram of an identity authentication device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0024] To make the purpose, technical solution and advantages of the embodiments of the present invention more clear, the embodiments of the present invention are further described in detail below in conjunction with the accompanying drawings. Here, the exemplary embodiments of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention.

[0025] The term "and / or" herein only describes an association relationship, indicating that three relationships may exist. For example, A and / or B may represent the following three situations: A exists alone, A and B exist at the same time, and B exists alone. In addition, the term "at least one" herein represents any combination of at least two of any one or more of a plurality of. For example, including at least one of A, B, and C may represent including any one or more elements selected from the set consisting of A, B, and C.

[0026] In the description of this specification, the terms "include", "including", "have", "contain", etc. are all open terms, which mean including but not limited to. The descriptions with reference to the terms "one embodiment", "a specific embodiment", "some embodiments", "for example", etc. mean that the specific features, structures or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures or characteristics described may be combined in a suitable manner in any one or more embodiments or examples. The order of steps involved in each embodiment is used to schematically illustrate the implementation of the present application, and the order of steps is not limited and can be appropriately adjusted as needed.

[0027] According to research, when a user goes to a bank to handle banking business, the user needs to use the ID card to authenticate the user. Only when the identity authentication is passed can the user handle the corresponding banking business. When the ID card is used to authenticate the user, the face image on the user's ID card is compared with the user's face image collected on site to see if they are consistent. If the comparison result is consistent, the user passes the identity authentication. If the comparison result is inconsistent, the user fails the identity authentication. This verification method, because it compares the face image on the user's ID card with the user's face image collected on site, in order to avoid the face image on the user's ID card and the face image collected on site are quite different, it will limit the ID card to be authenticated within the validity period. However, if the ID card is within the validity period but has not been replaced for a long time, the user's face image and the face image on the ID card may also be quite different. Therefore, the current identity authentication method only authenticates the user by comparing the face image on the user's ID card with the face image collected on site to see if they are consistent, resulting in inaccurate authentication results. In addition, when the ID card expires, the user cannot perform identity authentication. The identity authentication efficiency is low, which affects the efficiency of banking business and the user experience is poor.

[0028] In view of the above research, the present invention provides an identity authentication method, such as Figure 1 As shown, including:

[0029] S101: Collecting the current face image of the user and the face image on the user's ID card;

[0030] S102: extracting a first feature vector of the current face image and a second feature vector of the face image on the user's ID card using a pre-trained deep convolutional neural network; wherein the deep convolutional neural network is used to filter the input image using at least one convolutional layer to obtain a feature vector of the image;

[0031] S103: Calculating the cosine similarity between the current face image of the user and the face image on the user's ID card based on the first feature vector and the second feature vector;

[0032] S104: Input the banking service that the user wants to handle, the cosine similarity, and the user's identity information into a pre-trained GA-BP neural network to obtain a verification result of whether the user can handle the banking service; wherein the GA-BP neural network is used to authenticate the user's identity and determine whether the user can handle the input banking service.

[0033] In an embodiment of the present invention, a current facial image of a user and a facial image on the user's ID card are collected; a first feature vector of the current facial image and a second feature vector of the facial image on the user's ID card are extracted using a pre-trained deep convolutional neural network; wherein the deep convolutional neural network is used to filter an input image using at least one convolutional layer to obtain a feature vector of the image; the cosine similarity between the current facial image of the user and the facial image on the user's ID card is calculated based on the first feature vector and the second feature vector; the banking business to be handled by the user, the cosine similarity, and the user's identity information are input into the pre-trained GA-BP neural network to obtain a verification result of whether the user can handle the banking business; wherein the GA-BP neural network is used to authenticate the user's identity and determine whether the user can handle the input banking business. In this way, the GA-BP neural network can be used to obtain the verification result of whether the user can handle the bank business according to the user's identity information, cosine similarity, and the bank business to be handled. It is not only judged by the similarity between the face image on the user's ID card and the user's current face image. Therefore, the user can also verify whether he can handle the bank business when the ID card is expired. And when there is a large difference between the face image on the ID card and the user's current face image, it will not directly lead to the failure of the user's identity authentication and the inability to handle the bank business to be handled. The accuracy and efficiency of identity authentication are improved, thereby improving the efficiency of bank business handling and improving user experience.

[0034] The above S101 to S104 are described in detail below.

[0035] For the above S101, for example, the current facial image of the user can be collected by any of the following methods but not limited to: using an image acquisition device to collect a facial video of the user, and using frame extraction technology to extract a facial image from the facial video; or directly using an image acquisition device to capture the current facial image of the user.

[0036] In addition, the user ID card includes, for example: an ID card within the validity period and an ID card that is not within the validity period (ie, an expired ID card).

[0037] For the above S102, the deep convolutional neural network, for example, includes: an input layer, at least one convolutional layer, and an output layer; for example, the following method can be used but not limited to use a pre-trained deep convolutional neural network to extract the first feature vector of the current face image and the second feature vector of the face image on the user's ID card: the current face image is input through the input layer, and is sequentially processed by convolution of at least one convolutional layer, and the convolution result output by the last convolutional layer is input into the output layer, and the output layer uses a classifier to classify the convolution result to obtain the first feature vector of the current face image and output it; the face image on the user's ID card is input through the input layer, and is sequentially processed by convolution of at least one convolutional layer, and the convolution result output by the last convolutional layer is input into the output layer, and the output layer uses a classifier to classify the convolution result to obtain the second feature vector of the face image on the user's ID card and output it.

[0038] For example, Figure 2 As shown, it is a schematic diagram of the structure of a deep convolutional neural network provided by an embodiment of the present invention, including an input layer, four convolutional layers (a first convolutional layer, a second convolutional layer, a third convolutional layer, and a fourth convolutional layer), and an output layer. Taking the extraction of the first feature vector of the current face image of the user as an example, the current face image of the user is input into the first convolutional layer using the input layer, and the first convolutional layer includes, for example, 20 first convolutional kernels of size 4×4, and the current face image of the user is convolved with the 20 first convolutional kernels respectively, and then the results of the convolutional processing of the current face image by the 20 first convolutional kernels are added in a bitwise manner, and the first convolution result is output; the first convolution result is input into the second convolutional layer, and the second convolutional layer includes, for example, 40 second convolutional kernels of size 3×3, and the first convolution result is convolved with the 40 second convolutional kernels respectively, and then the result of the convolution processing of the first convolution result by the 20 second convolutional kernels is output. Add the bits in place and output the second convolution result; input the second convolution result to the third convolution layer, the third convolution layer for example includes 60 third convolution kernels with a convolution kernel size of 4×4, use the 60 third convolution kernels to perform convolution processing on the second convolution result respectively, then add the results of the convolution processing of the second convolution result by the 60 third convolution kernels in a positional manner, and output the third convolution result; input the third convolution result to the fourth convolution layer, the fourth convolution layer for example includes 60 fourth convolution kernels with a convolution kernel size of 3×3, use the 60 fourth convolution kernels to perform convolution processing on the third convolution result respectively, then add the results of the convolution processing of the third convolution result by the 60 fourth convolution kernels in a positional manner, and output the fourth convolution result; input the fourth convolution result to the output layer, the output layer for example includes a K-class softmax classifier, use the classifier to map each eigenvalue contained in the fourth convolution result to a preset interval, and output the first eigenvector.

[0039] Here, K is the number of eigenvalues ​​contained in the first eigenvector.

[0040] Specifically, the number of convolutional layers, the number of convolutional kernels contained in each convolutional layer, the size of each convolutional kernel, and the parameter values ​​contained in each convolutional kernel can be obtained by, for example, training a deep convolutional neural network using a face image with pre-labeled feature vectors as a training sample. In different application environments, different training samples can be used to train a deep neural network to obtain the number of convolutional layers, the number of convolutional kernels contained in each convolutional layer, the size of each convolutional kernel, and the parameter values ​​contained in each convolutional kernel that are suitable for the application environment. Therefore, the number of convolutional layers contained in a deep neural network, the number of convolutional kernels contained in each convolutional layer, and the size of each convolutional kernel and each convolutional kernel can be set in combination with specific scenarios, and there is no restriction here.

[0041] In addition, in order to further simplify the computational complexity of each convolution layer when performing convolution processing, in another embodiment of the present invention, the deep convolutional neural network, for example, further includes a pooling layer arranged between every two convolutional layers:

[0042] (1) Therefore, when the current face image is input through the input layer and sequentially subjected to convolution processing by at least one convolution layer, for example, the following method can be adopted: the current face image is input through the input layer, the current face image is processed by a convolution layer to output the convolution result, the convolution result is input to the pooling layer between the convolution layer and the next convolution layer for pooling processing, the pooling processing result is input to the next convolution layer for convolution processing, until the convolution result is output by the last convolution layer.

[0043] (2) Therefore, when the face image on the user's ID card is input through the input layer and sequentially subjected to convolution processing by at least one convolution layer, for example, the following method can be adopted: the face image on the user's ID card is input through the input layer, the face image on the user's ID card is processed by a convolution layer to output the convolution result, the convolution result is input to the pooling layer between the convolution layer and the next convolution layer for pooling processing, the pooling processing result is input to the next convolution layer for convolution processing, until the convolution result is output by the last convolution layer.

[0044] For example, Figure 3The figure is a schematic diagram of the structure of another deep convolutional neural network provided by an embodiment of the present invention, comprising an output layer, four convolutional layers (a first convolutional layer, a second convolutional layer, a third convolutional layer, and a fourth convolutional layer), three pooling layers (a first pooling layer, a second pooling layer, and a third pooling layer), and an output layer. The face image on the user's ID card is input through the input layer, and the convolution processing of at least one convolutional layer is performed in sequence. For example, the face image on the user's ID card is input into the first convolutional layer through the input layer, and the first convolutional layer performs convolution processing on the face image on the ID card to output a first convolution result, and the first convolution result is input into the first pooling layer. The pooling step size of the first pooling layer is, for example, 2, and the first pooling layer performs pooling processing on the first convolution result and outputs a first Pooling result; the first pooling result is input to the second convolution layer, the second convolution layer performs convolution processing on the first pooling result and outputs the second convolution result, the second convolution result is input to the second pooling layer, the pooling step size of the second pooling layer is, for example, 2, the second pooling layer performs pooling processing on the second convolution result and outputs the second pooling result, the second pooling result is input to the third convolution layer, the third convolution layer performs convolution processing on the second pooling result and outputs the third convolution result, the third convolution result is input to the third pooling layer, the pooling step size of the third pooling layer is, for example, 2, the third pooling layer performs pooling processing on the third convolution result and outputs the third pooling result; the third pooling result is input to the fourth convolution layer, the fourth convolution layer performs convolution processing on the third pooling result and outputs the fourth convolution result.

[0045] in, Figure 3 This is only an example of a deep neural network structure. The number of pooling layers in the deep neural network can be set in combination with the number of deep neural networks in actual application scenarios, and there is no restriction here.

[0046] For the above S103, for example, the cosine similarity between the current face image of the user and the face image on the user's ID card can be calculated based on the first feature vector and the second feature vector using, but not limited to, the following formula:

[0047]

[0048] Among them, cos(D i ,D j ) is the cosine similarity, D i is the first eigenvector, D j is the second eigenvector, W ik is the kth eigenvalue of the first eigenvector, W jk is the kth eigenvalue of the second eigenvector, n is the number of eigenvalues ​​included in the first eigenvector or the second eigenvector; the first eigenvector and the second eigenvector have the same number of eigenvalues.

[0049] With respect to the above S104, the identity information includes, for example, at least one of: ID number, mobile phone number, age, gender, name, etc.

[0050] Specifically, the GA-BP neural network includes, for example, an input layer, a hidden layer, and an output layer; Figure 4 As shown, it is a structural example diagram of a GA-BP neural network provided by an embodiment of the present invention. The input layer contains n nodes, the hidden layer contains p nodes, and the output layer contains m nodes, where n, p, and m are positive integers. Taking the first node of the output layer as an example, the output result of the output node is the weighted sum of the output results of each node in the hidden layer. For example, the hidden layer contains three nodes A, B, and C, and the output layer contains two nodes D and E. The weight of node A relative to node D is Q ad , the weight of node B relative to node D is Q bd , the weight of node C relative to node D is Q cd , the weight of node A relative to node E is Q ae , the weight of node B relative to node E is Q be , the weight of node C relative to node E is Q ce , then the output result of node D is: the output result of node A × Q ad + Output result of node B × Q bd + Output result of node C × Q cd ; When weighted summation is performed, the weight values ​​of each node in the hidden layer for different nodes in the output layer can be the same or different, for example, Q ad , Q ae Can be the same or different.

[0051] Here, the number of nodes in the output layer, the number of nodes in the output layer, the number of nodes in the hidden layer, the node parameters of each node in the hidden layer, and the weight value of each node in the hidden layer for each node in the output layer are determined by training using different training samples according to specific application scenarios, and are not restricted here.

[0052] like Figure 5 The flowchart of a method for training a GA-BP neural network provided by an embodiment of the present invention includes:

[0053] S501: Obtain training samples based on historical user identity information, cosine similarity, and historical banking transactions.

[0054] S502: Determine the number of nodes in the input layer of the GA-BP neural network according to the number of input information to be input, and determine the number of nodes in the output layer of the GA-BP neural network according to the output result to be obtained.

[0055] The input information includes, for example, user identity information, cosine similarity, historical banking transactions to be processed, etc.

[0056] Exemplarily, if the input information to be input is the user's name, ID number, the cosine similarity between the face image used for the ID card and the current face image, and the banking service to be handled, the number of nodes in the input layer is determined to be 4; the number of nodes in the output layer is determined based on the number of output results. For example, when the output result only includes the verification result of whether the user can handle the banking service, the output layer may only include one output node.

[0057] Alternatively, in another possible implementation, the output layer may include multiple nodes, and the verification result of whether the user can handle the banking business is obtained based on the ratio of the number of nodes that output the verification result that the user can handle the banking business to the total number of nodes in the output layer.

[0058] S503: Train the GA-BP neural network using the training samples to determine the number of nodes in the hidden layer of the GA-BP neural network.

[0059] Exemplarily, for example, the following method is used to train the GA-BP neural network using training samples to determine the number of nodes in the hidden layer of the GA-BP neural network: determine the initial number of nodes in a hidden layer; wherein the initial number of nodes can be, for example, a random value; loop through the following steps until the deviation between the first training loss and the test loss is not greater than a first preset threshold, and the first training loss and the test loss are less than a second preset threshold, determine the initial number of nodes in the current hidden layer as the number of nodes in the hidden layer, and stop looping:

[0060] Step ①: Divide the training samples into a test set and a training set; input the training samples in the training set into the GA-BP neural network to obtain the first output result.

[0061] Step ②: Calculate the first training loss based on the labels of the training samples in the training set and the first output result.

[0062] The label of the training sample includes a preset output result of the training sample, and the absolute value of the difference between the preset output result and the first output result is used as the first training loss.

[0063] Step ③: Input the training samples in the test set into the GA-BP neural network to obtain the second output result.

[0064] Step ④: Calculate the test loss according to the labels of the training samples in the test set and the second output result; if the deviation between the first training loss and the test loss is greater than the first preset threshold, reduce the number of nodes in the hidden layer to obtain a new initial number of nodes; if the deviation between the first training loss and the test loss is not greater than the first preset threshold, and the first training loss and the test loss are greater than the second preset threshold, increase the number of nodes in the hidden layer to obtain a new initial number of nodes.

[0065] Here, for example, the absolute value of the difference between the preset output result and the second output result is used as the test loss; the first preset threshold and the second preset threshold can be set in combination with specific scenarios, and are not limited here.

[0066] S504: Continue to train the GA-BP neural network using the training samples to determine the node parameters of each node in the hidden layer and the weight value of each node in the hidden layer corresponding to each node in the output layer.

[0067] Exemplarily, for example, the following method is used to continue to train the GA-BP neural network using training samples to determine the node parameters of each node in the hidden layer and the weight value of each node in the hidden layer corresponding to each node in the output layer: determine multiple initial value groups, wherein each initial value group contains the initial parameters of each node in the hidden layer and the initial weight value of each node in the hidden layer corresponding to each node in the output layer; loop the following steps until an initial value group with a second training loss less than a third preset threshold is obtained, the initial parameters of each node in the hidden layer in the initial value group with the smallest second training loss are determined as the node parameters of each node in the hidden layer, and the initial weight value of each hidden node in the initial value group with the smallest second training loss corresponding to each node in the output layer is determined as the weight value of each node in the hidden layer corresponding to each node in the output layer:

[0068] Step 1: respectively set up a GA-BP neural network for each initial value group, input the training sample into the GA-BP neural network, and obtain the third output result corresponding to each initial value group;

[0069] Step 2: Determine the second training loss corresponding to each initial value group according to the label of the training sample and the third output result corresponding to each initial value group;

[0070] For example, the absolute value of the difference between the preset output result and the third output result is used as the second training loss.

[0071] Step 3: When the second training loss of each initial value group is not less than the third preset threshold, at least one initial value group is selected from the multiple initial value groups to update the initial values ​​to obtain multiple updated initial value groups.

[0072] S505: Obtain a trained GA-BP neural network according to the number of nodes in the input layer, the number of nodes in the output layer, the number of nodes in the hidden layer, the node parameters of each node in the hidden layer, and the weight value of each node in the hidden layer corresponding to each node in the output layer.

[0073] The present invention also provides an identity authentication device in the embodiment as described in the following embodiment. Since the principle of solving the problem by the device is similar to that of the identity authentication method, the implementation of the device can refer to the implementation of the identity authentication method, and the repeated parts will not be repeated.

[0074] like Figure 6 As shown, it is a schematic diagram of an identity authentication device provided by an embodiment of the present invention, comprising: a collection module 601, an extraction module 602, a calculation module 603, and a processing module 604; wherein,

[0075] The acquisition module 601 is used to acquire the current face image of the user and the face image on the user's ID card;

[0076] An extraction module 602 is used to extract a first feature vector of the current face image and a second feature vector of the face image on the user's ID card using a pre-trained deep convolutional neural network; wherein the deep convolutional neural network is used to filter the input image using at least one convolutional layer to obtain a feature vector of the image;

[0077] A calculation module 603, configured to calculate the cosine similarity between the current face image of the user and the face image on the user's ID card according to the first feature vector and the second feature vector;

[0078] Processing module 604 is used to input the banking service to be handled by the user, cosine similarity, and the user's identity information into a pre-trained GA-BP neural network to obtain a verification result of whether the user can handle the banking service; wherein the GA-BP neural network is used to authenticate the user's identity and determine whether the user can handle the input banking service.

[0079] In one possible implementation, a deep convolutional neural network includes: an input layer, at least one convolutional layer, and an output layer; an extraction module, specifically used to input a current face image through the input layer, sequentially undergo convolution processing through at least one convolutional layer, input a convolution result output by the last convolutional layer into the output layer, the output layer uses a classifier to classify the convolution result, obtain a first feature vector of the current face image and output it; input a face image on a user's ID card through the input layer, sequentially undergo convolution processing through at least one convolutional layer, input a convolution result output by the last convolutional layer into the output layer, the output layer uses a classifier to classify the convolution result, obtain a second feature vector of the face image on the user's ID card and output it.

[0080] In one possible implementation, the deep convolutional neural network also includes: a pooling layer arranged between every two convolutional layers; an extraction module, which is specifically used to input the current face image through the input layer, process the current face image through a convolutional layer to output the convolution result, input the convolution result into the pooling layer between the convolutional layer and the next convolutional layer for pooling processing, input the pooling processing result into the next convolutional layer for convolution processing, until the convolution result output by the last convolutional layer; input the face image on the user's ID card through the input layer, process the face image on the user's ID card through a convolutional layer to output the convolution result, input the convolution result into the pooling layer between the convolutional layer and the next convolutional layer for pooling processing, input the pooling processing result into the next convolutional layer for convolution processing, until the convolution result output by the last convolutional layer.

[0081] In a possible implementation, the calculation module is specifically used to calculate the cosine similarity between the current face image of the user and the face image on the user's ID card according to the first feature vector and the second feature vector using the following formula:

[0082]

[0083] Among them, cos(D i ,D j ) is the cosine similarity, D i is the first eigenvector, D j is the second eigenvector, W ik is the kth eigenvalue of the first eigenvector, W jk is the kth eigenvalue of the second eigenvector, n is the number of eigenvalues ​​included in the first eigenvector or the second eigenvector; the first eigenvector and the second eigenvector have the same number of eigenvalues.

[0084] In a possible implementation, it also includes: a training module, which is used to obtain training samples based on historical user identity information, cosine similarity, and historical banking transactions; determine the number of nodes in the input layer of the GA-BP neural network based on the amount of input information to be input, and determine the number of nodes in the output layer of the GA-BP neural network based on the output results to be obtained; use the training samples to train the GA-BP neural network to determine the number of nodes in the hidden layer of the GA-BP neural network; continue to use the training samples to train the GA-BP neural network to determine the node parameters of each node in the hidden layer, and the weight value of each node in the hidden layer corresponding to each node in the output layer; obtain a trained GA-BP neural network based on the number of nodes in the input layer, the number of nodes in the output layer, the number of nodes in the hidden layer, the node parameters of each node in the hidden layer, and the weight value of each node in the hidden layer corresponding to each node in the output layer.

[0085] In a possible implementation, the training module is specifically used to determine the initial number of nodes in a hidden layer; looping the following steps until the deviation between the first training loss and the test loss is not greater than the first preset threshold, and the first training loss and the test loss are less than the second preset threshold, determining that the initial number of nodes in the current hidden layer is the number of nodes in the hidden layer, and stopping the looping steps: dividing the training samples into a test set and a training set; inputting the training samples in the training set into the GA-BP neural network to obtain a first output result; calculating the first training loss according to the labels of the training samples in the training set and the first output result; inputting the training samples in the test set into the GA-BP neural network to obtain a second output result; calculating the test loss according to the labels of the training samples in the test set and the second output result; if the deviation between the first training loss and the test loss is greater than the first preset threshold, reducing the number of nodes in the hidden layer to obtain a new initial number of nodes; if the deviation between the first training loss and the test loss is not greater than the first preset threshold, and the first training loss and the test loss are greater than the second preset threshold, increasing the number of nodes in the hidden layer to obtain a new initial number of nodes.

[0086] In a possible implementation, the training module is specifically used to determine multiple initial value groups, wherein each initial value group includes initial parameters of each node of the hidden layer and initial weight values ​​of each node of the hidden layer corresponding to each node of the output layer; the following steps are executed repeatedly until an initial value group with a second training loss less than a third preset threshold is obtained, the initial parameters of each node of the hidden layer in the initial value group with the smallest second training loss are determined as node parameters of each node of the hidden layer, and the initial weight values ​​of each hidden node corresponding to each node of the output layer in the initial value group with the smallest second training loss are determined as weight values ​​of each node of the hidden layer corresponding to each node of the output layer: a GA-BP neural network is set for each initial value group respectively, and a training sample is input into the GA-BP neural network to obtain a third output result corresponding to each initial value group; a second training loss corresponding to each initial value group is determined according to a label of the training sample and the third output result corresponding to each initial value group; when the second training loss of each initial value group is not less than the third preset threshold, at least one initial value group is selected from the multiple initial value groups to update each initial value, so as to obtain multiple updated initial value groups.

[0087] In a possible implementation, the user's identity information includes at least one of the following: ID number, mobile phone number, age, gender, and name.

[0088] An embodiment of the present invention further provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned identity authentication method when executing the computer program.

[0089] An embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program for executing the above-mentioned identity authentication method.

[0090] In an embodiment of the present invention, a current facial image of a user and a facial image on the user's ID card are collected; a first feature vector of the current facial image and a second feature vector of the facial image on the user's ID card are extracted using a pre-trained deep convolutional neural network; wherein the deep convolutional neural network is used to filter an input image using at least one convolutional layer to obtain a feature vector of the image; the cosine similarity between the current facial image of the user and the facial image on the user's ID card is calculated based on the first feature vector and the second feature vector; the banking business to be handled by the user, the cosine similarity, and the user's identity information are input into the pre-trained GA-BP neural network to obtain a verification result of whether the user can handle the banking business; wherein the GA-BP neural network is used to authenticate the user's identity and determine whether the user can handle the input banking business. In this way, the GA-BP neural network can be used to obtain the verification result of whether the user can handle the bank business according to the user's identity information, cosine similarity, and the bank business to be handled. It is not only judged by the similarity between the face image on the user's ID card and the user's current face image. Therefore, the user can also verify whether he can handle the bank business when the ID card is expired. And when there is a large difference between the face image on the ID card and the user's current face image, it will not directly lead to the failure of the user's identity authentication and the inability to handle the bank business to be handled. The accuracy and efficiency of identity authentication are improved, thereby improving the efficiency of bank business handling and improving user experience.

[0091] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0092] 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 flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, 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 flowchart and / or block diagram. 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.

[0093] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate 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 A function specified in one or more boxes.

[0094] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0095] 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. An identity authentication method, characterized in that: include: Collect the user's current face image and the face image on the user's ID card; Extracting a first feature vector of the current face image and a second feature vector of the face image on the user's ID card using a pre-trained deep convolutional neural network; wherein the deep convolutional neural network is used to filter the input image using at least one convolutional layer to obtain a feature vector of the image; Calculating the cosine similarity between the current face image of the user and the face image on the user's ID card based on the first feature vector and the second feature vector; Input the banking service to be handled by the user, the cosine similarity, and the user's identity information into the pre-trained GA-BP neural network to obtain a verification result of whether the user can handle the banking service; wherein the GA-BP neural network is used to authenticate the user's identity and determine whether the user can handle the input banking service; Also includes: The training samples are obtained based on the historical user's identity information, cosine similarity, and the historical banking transactions handled; Determine the number of nodes in the input layer of the GA-BP neural network according to the amount of input information to be input, and determine the number of nodes in the output layer of the GA-BP neural network according to the output results to be obtained; The GA-BP neural network is trained using training samples to determine the number of nodes in the hidden layer of the GA-BP neural network; Continue to train the GA-BP neural network using the training samples to determine the node parameters of each node in the hidden layer and the weight value of each node in the hidden layer corresponding to each node in the output layer; A trained GA-BP neural network is obtained according to the number of nodes in the input layer, the number of nodes in the output layer, the number of nodes in the hidden layer, the node parameters of each node in the hidden layer, and the weight value of each node in the hidden layer corresponding to each node in the output layer; The input information is the user's name, ID number, the cosine similarity between the face image of the user's ID card and the current face image, and the bank business to be handled. The number of nodes in the input layer is determined to be 4; the number of nodes in the output layer is determined according to the number of output results; Determining the number of nodes in the output layer according to the number of output results includes: when the output result only includes the verification result of whether the user can handle the banking business, the output layer only includes one output node; Alternatively, the output layer includes multiple nodes, and the verification result of whether the user can handle the banking business is obtained according to the ratio of the number of nodes that output the verification result that the user can handle the banking business to the total number of nodes in the output layer.

2. The identity authentication method according to claim 1, characterized in that: The deep convolutional neural network includes: an input layer, at least one convolutional layer, and an output layer; The first feature vector of the current face image and the second feature vector of the face image on the user's ID card are extracted using a pre-trained deep convolutional neural network, including: The current face image is input through the input layer, and is sequentially processed by at least one convolution layer. The convolution result output by the last convolution layer is input into the output layer. The output layer uses a classifier to classify the convolution result, obtains the first feature vector of the current face image, and outputs it. The face image on the user's ID card is input through the input layer, and is sequentially convolutionally processed by at least one convolution layer. The convolution result output by the last convolution layer is input into the output layer. The output layer uses a classifier to classify the convolution result, obtains the second feature vector of the face image on the user's ID card, and outputs it.

3. The identity authentication method according to claim 2, characterized in that: The deep convolutional neural network also includes: a pooling layer arranged between every two convolutional layers; The current face image is input through the input layer and sequentially subjected to convolution processing of at least one convolution layer, including: The current face image is input through the input layer, and after the current face image is processed by a convolution layer to output the convolution result, the convolution result is input to the pooling layer between the convolution layer and the next convolution layer for pooling processing, and the pooling processing result is input to the next convolution layer for convolution processing until the convolution result is output by the last convolution layer; The face image on the user's ID card is input through the input layer and sequentially subjected to convolution processing by at least one convolution layer, including: The face image on the user's ID card is input through the input layer. After the face image on the user's ID card is processed by a convolution layer to output the convolution result, the convolution result is input to the pooling layer between the convolution layer and the next convolution layer for pooling processing. The pooling processing result is input to the next convolution layer for convolution processing until the convolution result is output by the last convolution layer.

4. The identity authentication method according to claim 1, characterized in that: Calculating the cosine similarity between the current face image of the user and the face image on the user's ID card according to the first feature vector and the second feature vector includes: The cosine similarity between the current face image of the user and the face image on the user's ID card is calculated based on the first eigenvector and the second eigenvector using the following formula: Among them, cos(D i ,D j ) is the cosine similarity, D i is the first eigenvector, D j is the second eigenvector, W ik is the kth eigenvalue of the first eigenvector, W jk is the kth eigenvalue of the second eigenvector, n is the number of eigenvalues ​​included in the first eigenvector or the second eigenvector; the first eigenvector and the second eigenvector have the same number of eigenvalues.

5. The identity authentication method according to claim 1, characterized in that: The GA-BP neural network is trained using training samples to determine the number of nodes in the hidden layer of the GA-BP neural network, including: Determine the initial number of nodes in a hidden layer; The following steps are executed cyclically until the deviation between the first training loss and the test loss is not greater than the first preset threshold, and the first training loss and the test loss are less than the second preset threshold, the initial number of nodes of the current hidden layer is determined to be the number of nodes of the hidden layer, and the execution of the cyclic steps is stopped: Divide the training samples into test set and training set; Input the training samples in the training set into the GA-BP neural network to obtain the first output result; Calculate a first training loss according to the labels of the training samples in the training set and the first output result; Input the training samples in the test set into the GA-BP neural network to obtain the second output result; Calculate the test loss based on the labels of the training samples in the test set and the second output result; If the deviation between the first training loss and the test loss is greater than a first preset threshold, reducing the number of nodes in the hidden layer to obtain a new initial number of nodes; If the deviation between the first training loss and the test loss is not greater than a first preset threshold, and the first training loss and the test loss are greater than a second preset threshold, the number of nodes in the hidden layer is increased to obtain a new initial number of nodes.

6. The identity authentication method according to claim 1, characterized in that: Continue to train the GA-BP neural network using the training samples to determine the node parameters of each node in the hidden layer and the weight value of each node in the hidden layer corresponding to each node in the output layer, including: Determine a plurality of initial value groups, wherein each initial value group includes an initial parameter of each node of the hidden layer and an initial weight value of each node of the hidden layer corresponding to each node of the output layer; The following steps are executed in a loop until an initial value group with a second training loss less than a third preset threshold is obtained, the initial parameters of each node of the hidden layer in the initial value group with the smallest second training loss are determined as the node parameters of each node of the hidden layer, and the initial weight values ​​of each hidden node in the initial value group with the smallest second training loss corresponding to each node of the output layer are determined as the weight values ​​of each node of the hidden layer corresponding to each node of the output layer: A GA-BP neural network is set for each initial value group, and a training sample is input into the GA-BP neural network to obtain a third output result corresponding to each initial value group; Determine a second training loss corresponding to each initial value group according to the label of the training sample and the third output result corresponding to each initial value group; When the second training loss of each initial value group is not less than the third preset threshold, at least one initial value group is selected from the multiple initial value groups to update the initial values ​​to obtain multiple updated initial value groups.

7. The identity authentication method according to claim 1, characterized in that: The user's identity information includes at least one of the following: ID number, mobile phone number, age, gender, name.

8. An identity authentication device, characterized in that: include: The acquisition module is used to acquire the current face image of the user and the face image on the user's ID card; An extraction module, used to extract a first feature vector of the current face image and a second feature vector of the face image on the user's ID card using a pre-trained deep convolutional neural network; wherein the deep convolutional neural network is used to filter the input image using at least one convolutional layer to obtain a feature vector of the image; A calculation module, used for calculating the cosine similarity between the current face image of the user and the face image on the user's ID card according to the first feature vector and the second feature vector; The processing module is used to input the banking service to be handled by the user, the cosine similarity, and the user's identity information into the pre-trained GA-BP neural network to obtain a verification result of whether the user can handle the banking service; wherein the GA-BP neural network is used to authenticate the user's identity and determine whether the user can handle the input banking service; It also includes: a training module, which is used to obtain training samples based on historical user identity information, cosine similarity, and historical banking transactions handled; Determine the number of nodes in the input layer of the GA-BP neural network according to the amount of input information to be input, and determine the number of nodes in the output layer of the GA-BP neural network according to the output results to be obtained; The GA-BP neural network is trained using training samples to determine the number of nodes in the hidden layer of the GA-BP neural network; Continue to train the GA-BP neural network using the training samples to determine the node parameters of each node in the hidden layer and the weight value of each node in the hidden layer corresponding to each node in the output layer; A trained GA-BP neural network is obtained according to the number of nodes in the input layer, the number of nodes in the output layer, the number of nodes in the hidden layer, the node parameters of each node in the hidden layer, and the weight value of each node in the hidden layer corresponding to each node in the output layer; The input information is the user's name, ID number, the cosine similarity between the face image of the user's ID card and the current face image, and the bank business to be handled. The number of nodes in the input layer is determined to be 4; the number of nodes in the output layer is determined according to the number of output results; Determining the number of nodes in the output layer according to the number of output results includes: when the output result only includes the verification result of whether the user can handle the banking business, the output layer only includes one output node; Alternatively, the output layer includes multiple nodes, and the verification result of whether the user can handle the banking business is obtained according to the ratio of the number of nodes that output the verification result that the user can handle the banking business to the total number of nodes in the output layer.

9. The identity authentication device as claimed in claim 8, characterized in that: The training module is specifically used to determine the initial number of nodes in a hidden layer; The following steps are executed cyclically until the deviation between the first training loss and the test loss is not greater than the first preset threshold, and the first training loss and the test loss are less than the second preset threshold, the initial number of nodes of the current hidden layer is determined to be the number of nodes of the hidden layer, and the execution of the cyclic steps is stopped: Divide the training samples into test set and training set; Input the training samples in the training set into the GA-BP neural network to obtain the first output result; Calculate a first training loss according to the labels of the training samples in the training set and the first output result; Input the training samples in the test set into the GA-BP neural network to obtain the second output result; Calculate the test loss based on the labels of the training samples in the test set and the second output result; If the deviation between the first training loss and the test loss is greater than a first preset threshold, reducing the number of nodes in the hidden layer to obtain a new initial number of nodes; If the deviation between the first training loss and the test loss is not greater than a first preset threshold, and the first training loss and the test loss are greater than a second preset threshold, the number of nodes in the hidden layer is increased to obtain a new initial number of nodes.

10. The identity authentication device according to claim 8, characterized in that: The training module is specifically used to determine a plurality of initial value groups, wherein each initial value group includes an initial parameter of each node of the hidden layer and an initial weight value of each node of the hidden layer corresponding to each node of the output layer; The following steps are executed in a loop until an initial value group with a second training loss less than a third preset threshold is obtained, the initial parameters of each node of the hidden layer in the initial value group with the smallest second training loss are determined as the node parameters of each node of the hidden layer, and the initial weight values ​​of each hidden node in the initial value group with the smallest second training loss corresponding to each node of the output layer are determined as the weight values ​​of each node of the hidden layer corresponding to each node of the output layer: A GA-BP neural network is set for each initial value group, and a training sample is input into the GA-BP neural network to obtain a third output result corresponding to each initial value group; Determine a second training loss corresponding to each initial value group according to the label of the training sample and the third output result corresponding to each initial value group; When the second training loss of each initial value group is not less than the third preset threshold, at least one initial value group is selected from the multiple initial value groups to update the initial values ​​to obtain multiple updated initial value groups.

11. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

12. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program for executing the method according to any one of claims 1 to 7.