Self-service method, system and terminal for user biological feature recognition

Abstract

The invention relates to a self-service method, system and terminal for user biological feature recognition, and relates to the field of identity authentication, and the method comprises the steps: obtaining a face image; when there is shielding, performing statistics on a first identity recognition accuracy rate satisfying a shielding region type in the face recognition historical data; when the accuracy is less than or equal to the accuracy threshold, randomly deleting the occlusion region type to obtain an updated occlusion region type; counting a second identity recognition accuracy rate meeting the updated shielding region type in the face recognition historical data; and when the accuracy is greater than the accuracy threshold, repairing the face image according to the deletion region type identifier, and obtaining identity information after obtaining the repaired face image and carrying out face recognition on the repaired face image. The technical problem that in the prior art, for shielded face image recognition, due to the fact that an enhanced local feature recognition strategy is unstable and the requirement for the computing power of repairing the whole face image recognition strategy is large, a shielded face recognition scheme giving consideration to the face recognition accuracy and timeliness at the same time lacks exists is solved.

Description

Technical Field

[0001] The present invention relates to the field of identity authentication, and in particular to a self-service method, system and terminal for user biometric identification. Background Art

[0002] Facial recognition is widely used in self-service products. In actual applications, facial recognition requires taking images of the entire face. However, in special demand scenarios such as influenza, taking images of the entire face is somewhat difficult. At this time, it is necessary to recognize occluded facial images.

[0003] There are two main methods for occluded image recognition. One is to repair the entire facial image, but the computing power required to repair the entire facial image is large and cannot meet the timeliness requirements. The other is to enhance the recognition of local features. Because the type of occluded area cannot be predicted, the disadvantage is that the stability is weak. Therefore, how to improve the accuracy of occluded face recognition while ensuring timeliness has become a technical problem that needs to be solved urgently. Summary of the invention

[0004] The present invention aims to solve the technical problem in the prior art of occluded face image recognition, which is caused by the instability of the enhanced local feature recognition strategy and the large computing power requirement of the entire facial image recognition strategy, resulting in a lack of an occluded face recognition solution that takes into account both face recognition accuracy and timeliness. A self-service method, system and terminal for user biometric recognition are provided to solve the problem.

[0005] The technical solution of the present invention to solve the above technical problems is as follows: In a first aspect, the present invention provides a self-service method for user biometric identification, comprising: obtaining a facial image; when the facial image is occluded, counting the first identity recognition accuracy rate that satisfies the occluded area type in the facial recognition history data; when the first identity recognition accuracy rate is less than or equal to an accuracy threshold, randomly deleting the occluded area type to obtain an updated occluded area type, wherein the updated occluded area type has a deleted area type identifier; counting the second identity recognition accuracy rate that satisfies the updated occluded area type in the facial recognition history data; when the second identity recognition accuracy rate is greater than the accuracy threshold, repairing the facial image according to the deleted area type identifier, obtaining the repaired facial image for facial recognition, and obtaining identity information.

[0006] In a second aspect, the present invention provides a self-service system for user biometric identification, comprising: a face image acquisition module, used to obtain a face image; a first recognition accuracy analysis module, used to count the first identity recognition accuracy that meets the occlusion area type in the face recognition history data when the face image is occluded; an occlusion area type deletion module, used to randomly delete the occlusion area type when the first identity recognition accuracy is less than or equal to an accuracy threshold, and obtain an updated occlusion area type, wherein the updated occlusion area type has a deletion area type identifier; a second recognition accuracy analysis module, used to count the second identity recognition accuracy that meets the updated occlusion area type in the face recognition history data; a biometric identification module, used to repair the face image according to the deletion area type identifier when the second identity recognition accuracy is greater than the accuracy threshold, and obtain identity information after face recognition is performed on the repaired face image.

[0007] In a third aspect, the present invention provides a self-service terminal for user biometric identification, which is used to implement a self-service method for user biometric identification as described in any one of the first aspects.

[0008] The beneficial effects of the present invention are as follows: when a face image is occluded, the current occluded area type is analyzed to statistically determine the identity recognition accuracy; when the identity recognition accuracy is too low, the occluded area type is randomly deleted to obtain an updated occluded area type; when the second identity recognition accuracy of the updated occluded area type is greater than the accuracy threshold, the face image is repaired according to the deleted area type identifier to obtain a repaired face image for face recognition; the area type that needs to be repaired is selected according to the identity recognition accuracy, and then face recognition is performed without repairing the entire facial image, thereby reducing computing power requirements and improving processing efficiency; at the same time, due to the supplementary repair features, the recognition accuracy is guaranteed, thereby achieving a technical effect of both accuracy and timeliness. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 A flowchart of a self-service method for user biometric identification provided by the present invention; Figure 2 A schematic diagram of the structure of a self-service system for user biometric identification provided by the present invention.

[0010] In the accompanying drawings, the components represented by each reference numeral are described as follows: face image acquisition module 100, first recognition accuracy analysis module 200, occlusion area type deletion module 300, second recognition accuracy analysis module 400, and biometric feature recognition module 500. DETAILED DESCRIPTION

[0011] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0012] In the description of the present invention, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0013] In the description of the present invention, the term "for example" is used to mean "used as an example, illustration or explanation". Any embodiment described as "for example" in the present invention is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is given to enable any technician in the field to implement and use the present invention. In the following description, details are listed for the purpose of explanation. It should be understood that a person of ordinary skill in the art can recognize that the present invention can be implemented without using these specific details. In other examples, well-known structures and processes will not be elaborated in detail to avoid obscuring the description of the present invention with unnecessary details. Therefore, the present invention is not intended to be limited to the embodiments shown, but is consistent with the widest scope consistent with the principles and features disclosed in the present invention.

[0014] Embodiment 1: like Figure 1 As shown, the embodiment of the present invention provides a self-service method for user biometric identification, including the steps of: Specifically, the self-service method for user biometric identification provided in the embodiment of the present application can be embedded in any self-service terminal, for example, such as a self-service store, a self-service cafeteria, etc. By implementing the self-service method for user biometric identification provided in the embodiment of the present application, the accuracy of face recognition in occluded scenes can be improved, and the intelligence and timeliness of self-service can be improved. Preferably, the self-service terminal can be deployed on a local device or on the cloud, and specifically includes a memory and a processor. The memory is used to store computer instructions corresponding to any step of the self-service method for user biometric identification provided in the embodiment of the present application, and the processor is used to retrieve the computer instruction execution process corresponding to any step of the self-service method for user biometric identification provided in the embodiment of the present application from the memory.

[0015] S10: obtaining a face image; Specifically, the facial image refers to the image data of the user's face captured by the image acquisition device (such as a camera) of the self-service terminal; the acquisition of the facial image is the starting point of the entire recognition process. The self-service terminal acquires the user's facial image through a built-in or external camera. Preferably, the image acquisition device has a self-checking function. When the acquired facial image is blurred or the optical fiber is dark, an error message will be fed back to prompt the user to re-enter the facial image to ensure the clarity of the acquired image.

[0016] S20: when the face image is blocked, counting the first identity recognition accuracy rate that meets the blocked area type in the face recognition history data; Specifically, occlusion in a facial image means that part of the facial image is blocked by objects (such as masks, glasses, hats, etc.), resulting in the inability to fully present the entire face; the occlusion area type refers to the specific occluded part (such as mouth, nose, eyes, etc.) and its shape and range; facial recognition historical data refers to the historical record data of facial recognition, including occlusion area type record data before facial recognition, and a facial recognition success record flag. When the facial recognition success record flag is 1, the identity recognition is accurate; when the facial recognition success record flag is 0, the identity recognition is incorrect. Preferably, the facial recognition historical data can collect historical data within three years; the first identity recognition accuracy rate refers to the proportion of successful identification of correct identities in the facial recognition historical data when identifying similar facial images with the aforementioned occlusion area types.

[0017] By analyzing the recognition performance under the same occlusion type in historical data, a decision-making basis is provided for subsequent repair strategies.

[0018] S30: When the first identity recognition accuracy is less than or equal to an accuracy threshold, randomly delete the occlusion area type to obtain an updated occlusion area type, wherein the updated occlusion area type has a deleted area type identifier; Specifically, the accuracy threshold refers to a preset minimum acceptable identity recognition accuracy standard, which is used to determine whether the recognition result under the current occlusion situation is reliable. Users can customize the identity recognition accuracy through the self-service terminal. If no customized configuration is made, the accuracy threshold defaults to 95%; updating the occlusion area type refers to the new occlusion area type obtained after randomly deleting the occlusion area type. Preferably, the number of randomly deleted types is at least 1 and the maximum does not exceed the total number of occlusion area types; deleting the area type identifier refers to the area type identifier selected for deletion during random deletion.

[0019] When the accuracy of the first identity recognition is lower than or equal to the accuracy threshold, it means that the current occlusion situation interferes too much with face recognition and the reliability of direct recognition is low; at this time, some occlusion area types will be randomly deleted to generate a new occlusion area combination (i.e., the occlusion area type is updated).

[0020] By reducing the number of occluded areas, we try to reduce the negative impact of occlusion on recognition; at the same time, we will record the type of deleted occluded areas and add a deleted area type identifier to them so that subsequent steps can repair these areas in a targeted manner.

[0021] S40: Counting the second identity recognition accuracy rate that meets the update occlusion area type in the face recognition history data; Specifically, when the updated occlusion area type is obtained, the face recognition history data will be retrieved again to find the historical records that match the updated occlusion area type, and the accuracy of its identity recognition will be counted.

[0022] By analyzing the recognition performance under the same updated occlusion type in historical data, a decision-making basis is provided for subsequent repair strategies.

[0023] S50: When the second identity recognition accuracy is greater than the accuracy threshold, the face image is repaired according to the deleted area type identifier, and the repaired face image is obtained for face recognition to obtain identity information.

[0024] Specifically, when the second identity recognition accuracy of the updated occluded area type is higher than the accuracy threshold, the face image will be repaired according to the deleted area type identifier.

[0025] The specific process is as follows: first, the pre-trained face image repairer is called, and the deleted occluded area in the face image is repaired in a targeted manner based on the deleted area type identifier.

[0026] The repaired image is closer to a complete face image, and then the repaired face image is input into the face recognition module for recognition, and finally the user's identity information is obtained.

[0027] By repairing the deleted occluded areas, the recognition accuracy can be further improved while ensuring processing efficiency, avoiding the need to repair the entire image, thereby significantly reducing computing power requirements. At the same time, by supplementing the repaired features, the reliability of the recognition results is ensured. Ultimately, the system can complete face recognition tasks efficiently and accurately, solving the technical problem of balancing accuracy and timeliness in the case of occlusion.

[0028] Further, the accuracy rate of first identity recognition that meets the occlusion area type in the historical face recognition data is counted, and step S20 includes the following steps: S21: the face recognition history data has an occlusion area record type; S22: extracting the occluded face recognition historical data whose occluded area record type overlaps with the occluded area type from the face recognition historical data, and counting the proportion of data with accurate identity recognition, which is set as the first identity recognition accuracy rate.

[0029] In order to count the accuracy of the first identity recognition, we first need to extract records that match the current occlusion area type from the historical face recognition data. The specific process is as follows: traverse the historical face recognition data and find out the records in which the occlusion area record type coincides with the current occlusion area type, that is, find out the cases in the historical data that have the same occlusion situation as the current one; then, count the amount of data in which the correct identity is successfully identified in these similar cases, and calculate its proportion of the total data volume, to get the first identity recognition accuracy.

[0030] This step plays a key evaluation role in the entire solution. By counting the first identity recognition accuracy, the impact of the current occlusion situation on face recognition can be evaluated. If the first identity recognition accuracy is lower than the preset threshold, it means that the current occlusion situation has a greater impact on recognition and requires further processing (such as randomly deleting the occlusion area type). If the first identity recognition accuracy is greater than the preset threshold, it means that the current occlusion situation has little impact on recognition, and the face recognition module can be directly called for recognition to ultimately obtain the user's identity information.

[0031] Providing the first identity recognition accuracy provides a decision-making basis for subsequent repair strategies, ensuring the accuracy and efficiency of the entire recognition process.

[0032] Furthermore, step S30 also includes the steps of: S31: the updating of the occlusion area type is obtained by randomly deleting k types of the occlusion area type, wherein the number of occlusion area types is ≥ k ≥ 1, and the initial value of k is equal to 1; S32: When the second identity recognition accuracy is less than the accuracy threshold, update the occlusion area type according to the k value.

[0033] Specifically, random deletion of k types refers to randomly selecting k occlusion areas from the current occlusion area type for deletion operation to generate a new occlusion area combination, where k is a variable parameter indicating the number of deleted occlusion areas; the number of occlusion area types ≥ k ≥ 1 indicates the value range of k, that is, the number of occlusion areas deleted each time is at least 1 and does not exceed the total number of current occlusion areas; the initial value of k is equal to 1, indicating that when the random deletion operation is performed for the first time, the initial value of k is 1, that is, an attempt is made to delete an occlusion area first; updating the updated occlusion area type according to the k value means that when the second identity recognition accuracy corresponding to the updated occlusion area type (that is, the type after deleting k occlusion areas) is still lower than the accuracy threshold, it is necessary to further adjust the occlusion area type according to the k value, and then count the identity recognition accuracy again until the occlusion area deletion result that meets the accuracy threshold is obtained, so as to stop and try to improve the recognition accuracy.

[0034] The process of updating the occlusion area type is a dynamic and step-by-step optimization process. When the first identity recognition accuracy is lower than the threshold, we first try to randomly delete an occlusion area (k=1) to generate a new occlusion area combination. Then we calculate the second identity recognition accuracy of this combination.

[0035] If the second identity recognition accuracy is still lower than the threshold, the strategy will be further adjusted according to the value of k: 1. If the occlusion area combination corresponding to the current k value still does not achieve the expected recognition accuracy, try to increase the value of k (for example, from 1 to 2), that is, delete more occlusion areas to generate a new occlusion area combination.

[0036] 2. Repeat the above process, gradually increasing the value of k, until a combination of occluded areas is found that can make the identity recognition accuracy higher than the threshold.

[0037] By gradually increasing the number of deleted occluded areas, the system can dynamically adjust the combination of occluded areas to find the optimal occluded area type, thereby gradually improving the accuracy of face recognition while ensuring processing efficiency, avoiding blindly repairing the entire image and achieving a balance between accuracy and efficiency through targeted adjustments.

[0038] Further, the occlusion region type is updated according to the k value, and step S32 includes the following steps: S321: according to the k value, enumerate the types of the occluded regions in combination to obtain a plurality of type combinations; S322: When at least one of the plurality of type combinations has not been randomly deleted, updating the updated occluded area type according to the type combination that has not been randomly deleted; S323: When the plurality of type combinations have been randomly deleted, the k value is increased by one and the occlusion area type is updated.

[0039] Specifically, combination enumeration refers to the permutation and combination of occlusion area types to generate all possible occlusion area combinations. For example, if the occlusion area types include mouth, nose and eyes, the combination enumeration will generate all possible combinations such as "mouth + nose", "mouth + eyes", "nose + eyes", etc.; not randomly deleted means that some occlusion area combinations have not been tried in the previous random deletion operation; k value plus one means that when the current occlusion area combinations have been tried and the target accuracy has not been achieved, the number of randomly deleted occlusion areas is increased (that is, the value of k is increased) to try new combinations.

[0040] By gradually enumerating and trying different combinations of occlusion areas, we can systematically find the optimal occlusion area type, thereby gradually improving the accuracy of face recognition while ensuring processing efficiency, avoiding blind attempts. Instead, through orderly combination enumeration and gradually increasing the k value, we ensure that the best solution is found as much as possible under limited computing resources.

[0041] Further, when the face image is blocked, step S20 includes the steps of: S201: Obtain the type of the occluding object and the type of the occluding constraint area; S202: Retrieving the occlusion object type and the occlusion constraint area type through an encoder to perform occlusion encoding on a preset face image, and obtaining an occlusion image set and an occlusion area type identifier set; S203: using the occlusion region type identification set as supervision data and the occlusion image set as input data to train the occlusion classifier; S204: Processing the face image through the occlusion classifier to obtain the occlusion area type; S205: When the occlusion region type is empty, the face image has no occlusion; S206: When the occlusion region type is not empty, the face image has occlusion.

[0042] Specifically, the present application preferably processes the facial image through an occlusion classifier to determine whether there is occlusion, processes the facial image through the occlusion classifier, and obtains the occlusion area type; when the occlusion area type is empty, the facial image has no occlusion; when the occlusion area type is not empty, the facial image has occlusion.

[0043] In detail, the steps for constructing the occlusion classifier are as follows: the user can determine the possible types of occluding objects (masks, hats, glasses, etc.) and the corresponding occluding area ranges, that is, the occluding constraint area types, according to the actual application scenarios; the preset face images are processed by the encoder to simulate the combination of different occluding objects and occluding areas, and an occluding image set and an occluding area type identification set are generated. The encoder can solve the problems of insufficient training data and difficulty in identification of the occlusion classifier. Preferably, the encoder is based on generative adversarial neural network training. During the specific training, the user will first configure a batch of actually collected face images with occlusion, with the identification data being 1, and manually identify the unoccluded faces. A set of face images with occlusions obtained by re-picking the face image is identified as 0, and a binary classification verification component is trained. After the binary classification verification component converges, the preset face image, occluded object type, and occluded constraint area type are used as input to output a coded occluded image, which is input into the binary classification verification component. When the verification result of the binary classification verification group is 1 for at least 480 times of the coded image for 500 consecutive times, the encoder is considered to be trained successfully. At this time, the binary classification verification component is deleted and the encoder is stored. Further, the occluded area type identification set is used as supervision data and the occluded image set is used as input data to train the occlusion classifier so that it can accurately identify the occluded area type.

[0044] The pre-trained occlusion classifier can quickly and accurately detect the type of occluded area in the face image, providing a basis for subsequent accuracy statistics and repair strategies. This process ensures that the system can effectively identify occlusion situations, thereby achieving efficient and accurate recognition of occluded faces.

[0045] Further, the face image is repaired according to the deleted area type identifier to obtain a repaired face image, and step S50 includes: Taking the preset face image as supervision and taking the occlusion region type identification set and the occlusion image set as input, training a face image repairer; According to the face image restorer, the deletion region type identifier and the face image are processed to obtain the restored face image.

[0046] Specifically, the purpose of face image restoration is to supplement or correct the occluded area in the face image through the repairer, so as to improve the accuracy of face recognition. The specific process is as follows: Training a face image repairer: Using a preset face image (unobstructed complete face image) as supervision data, and an occluded region type identification set and an occluded image set as input data, the face image repairer is trained. By learning these data, the face image repairer learns how to restore the face image to a state close to that of being unobstructed based on the features and position information of the occluded region. Specifically, the occluded region type identification set and the occluded image set edited by the aforementioned encoder are used as input, the preset face image is used as supervision, and the inverse of the face image similarity is used as the loss function. When the loss function output value for a preset number of consecutive times (at least 500 times) is less than the loss function output threshold, a face image repairer is generated. The encoded data is used, the manually labeled data is omitted, and the intelligence of the model training is improved.

[0047] Repairing face images: According to the deleted area type identifier, the repairer processes the occluded area in the face image. The repairer will fill or correct the deleted occluded area according to the learned features and patterns, and finally generate a repaired face image.

[0048] By supplementing the occluded area with the repairer, the facial image can be closer to a complete state, thereby improving the accuracy of face recognition. It not only avoids repairing the entire image (thereby reducing computing power requirements), but also ensures the accuracy and reliability of the repair effect through targeted repair. Ultimately, the repaired facial image can better support the face recognition module and achieve efficient and accurate identity recognition.

[0049] Furthermore, face recognition is performed, including: Obtain a set of verification face images corresponding to the entered identity database; Traversing the verification face image set, comparing it with the repaired face image, and obtaining a verification similarity set; The input identity of the verification face image with the maximum verification similarity is extracted from the input identity database and set as the identity information.

[0050] Specifically, the entry identity database refers to a database that stores user identity information and its corresponding facial images, which is used for identity authentication; the verification facial image set refers to the set of all facial images associated with the identity to be identified in the entry identity database, and these images are used to compare with the restored facial image; the verification similarity set refers to the set of similarity values ​​obtained after comparing the restored facial image with each image in the verification facial image set; the entry identity refers to the user identity information corresponding to the maximum verification similarity value, that is, the identity finally identified.

[0051] The verification similarity evaluation module of the embodiment of the present application preferably adopts the face recognition module constructed based on the twin neural network provided in the embodiment of the present application to perform face recognition.

[0052] In detail, the steps for building a face recognition module based on a twin neural network are as follows: A feature extraction channel is configured based on a convolutional neural network to extract texture features, shape features and color features of the image; then a first weight, a second weight and a third weight are configured for the texture features, the shape features and the color features, respectively, wherein the second weight is greater than the sum of the first weight and the third weight; then a similarity calculation channel is configured to calculate the Euclidean distance of the texture feature, the Euclidean distance of the shape feature and the Euclidean distance of the color feature respectively; the reciprocal of the Euclidean distance of the texture feature, the reciprocal of the Euclidean distance of the shape feature and the reciprocal of the Euclidean distance of the color feature are weightedly summed according to the first weight, the second weight and the third weight to obtain a similarity calculation value.

[0053] The specific implementation process is: The verification face image is input into the feature extraction channel to extract the verification texture features, verification shape features and verification color features; then the repaired face image is input into the feature extraction channel to extract the texture features to be analyzed, the shape features to be analyzed and the color features to be analyzed; then through the similarity calculation channel, the verification texture features, verification shape features and verification color features are compared with the texture features to be analyzed, the shape features to be analyzed and the color features to be analyzed, and the verification similarity is output and added to the verification similarity set for subsequent identity verification.

[0054] By comparing with the verified face image entered into the identity database, the user identity corresponding to the repaired face image can be determined. At the same time, combined with the previous occlusion detection and image restoration steps, efficient and accurate face recognition in occluded conditions can be achieved.

[0055] The self-service method for user biometric identification provided by the embodiment of the present invention has at least the following technical effects: 1. When the face image is occluded, the current occluded area type is analyzed to calculate the identity recognition accuracy. When the identity recognition accuracy is too low, the occluded area type is randomly deleted to obtain an updated occluded area type. When the second identity recognition accuracy of the updated occluded area type is greater than the accuracy threshold, the face image is repaired according to the deleted area type identifier to obtain a repaired face image for face recognition. The area type that needs to be repaired is selected according to the identity recognition accuracy, and then face recognition is performed. There is no need to repair the entire facial image, which reduces computing power requirements and improves processing efficiency. At the same time, due to the supplementary repair features, the recognition accuracy is guaranteed, thereby achieving a technical effect that takes into account both accuracy and timeliness.

[0056] 2. By gradually increasing the number of deleted occlusion areas, the system can dynamically adjust the combination of occlusion areas to find the optimal occlusion area type, thereby gradually improving the accuracy of face recognition while ensuring processing efficiency, avoiding blindly repairing the entire image, and achieving a balance between accuracy and efficiency through targeted adjustments.

[0057] 3. By gradually enumerating and trying different combinations of occlusion areas, the optimal occlusion area type can be systematically found, thereby gradually improving the accuracy of face recognition while ensuring processing efficiency, avoiding blind attempts, and instead ensuring that the optimal solution is found as much as possible under limited computing resources through orderly combination enumeration and gradual increase of k values.

[0058] Embodiment 2: like Figure 2 As shown, based on the same inventive concept as the self-service method for user biometric identification provided in the first embodiment, the embodiment of the present invention further provides a self-service system for user biometric identification, including: A face image acquisition module 100, used to obtain a face image; The first recognition accuracy analysis module 200 is used to count the first identity recognition accuracy rate that meets the occlusion area type in the face recognition history data when the face image is occluded; The occlusion region type deletion module 300 is used to randomly delete the occlusion region type when the first identity recognition accuracy is less than or equal to the accuracy threshold, and obtain an updated occlusion region type, wherein the updated occlusion region type has a deleted region type identifier; A second recognition accuracy analysis module 400, used to count the second identity recognition accuracy that meets the update occlusion area type in the face recognition history data; The biometric feature recognition module 500 is used to repair the face image according to the deleted area type identifier when the second identity recognition accuracy is greater than the accuracy threshold, obtain the repaired face image, perform face recognition, and obtain identity information.

[0059] Furthermore, the first recognition accuracy analysis module 200 executes the following steps: The face recognition history data has an occlusion area record type; The occluded face recognition historical data whose occluded area record type overlaps with the occluded area type is extracted from the face recognition historical data, and the proportion of data with accurate identity recognition is counted, which is set as the first identity recognition accuracy rate.

[0060] Furthermore, the occlusion region type deletion module 300 executes the following steps: The updating of the occlusion area type is obtained by randomly deleting k types of the occlusion area type, wherein the number of occlusion area types is ≥ k ≥ 1, and the initial value of k is equal to 1; When the second identity recognition accuracy is less than the accuracy threshold, the occlusion area type is updated according to the k value.

[0061] Furthermore, the occlusion region type deletion module 300 executes the following steps: According to the k value, the types of the occluded areas are combined and enumerated to obtain a plurality of type combinations; When at least one of the plurality of type combinations has not been randomly deleted, updating the updated occluded area type according to the type combination that has not been randomly deleted; When the plurality of type combinations have all been randomly deleted, the k value is increased by one and the occluded area type is updated.

[0062] Furthermore, the first recognition accuracy analysis module 200 executes the following steps: Get the occlusion object type and occlusion constraint area type; Retrieving the occlusion object type and the occlusion constraint area type through an encoder to perform occlusion encoding on a preset face image, and obtaining an occlusion image set and an occlusion area type identifier set; Using the occlusion region type identification set as supervision data and the occlusion image set as input data, training the occlusion classifier; Processing the face image through the occlusion classifier to obtain the occlusion area type; When the occlusion area type is empty, the face image has no occlusion; When the occlusion region type is not empty, the face image has occlusion.

[0063] Furthermore, the biometric identification module 500 performs the following steps: Taking the preset face image as supervision and taking the occlusion region type identification set and the occlusion image set as input, training a face image repairer; According to the face image restorer, the deletion region type identifier and the face image are processed to obtain the restored face image.

[0064] Furthermore, the biometric identification module 500 performs the following steps: Obtain a set of verification face images corresponding to the entered identity database; Traversing the verification face image set, comparing it with the repaired face image, and obtaining a verification similarity set; The input identity of the verification face image with the maximum verification similarity is extracted from the input identity database and set as the identity information.

[0065] Embodiment three: Based on the same inventive concept as the self-service method for user biometric identification provided in Example 1, an embodiment of the present invention further provides a self-service terminal for user biometric identification, which is used in the self-service method for user biometric identification described in any one of Example 1.

[0066] 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 computer, 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.

[0067] 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.

[0068] 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.

[0069] Although preferred embodiments of the present invention have been described, additional changes and modifications may occur to these embodiments once those skilled in the art understand the basic inventive concepts.

[0070] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention belong to the scope of the present invention and its equivalent technologies, the present invention is also intended to include these changes and variations.

Claims

1. A self-service method for user biometric identification, characterized in that: include: Obtaining a face image; When the face image is blocked, the first identity recognition accuracy rate that meets the blocked area type in the face recognition history data is counted; When the first identity recognition accuracy is less than or equal to the accuracy threshold, the occlusion area type is randomly deleted to obtain an updated occlusion area type, wherein the updated occlusion area type has a deleted area type identifier; Counting the second identity recognition accuracy rate that meets the update occlusion area type in the historical face recognition data; When the second identity recognition accuracy is greater than the accuracy threshold, the face image is repaired according to the deleted area type identifier, and the identity information is obtained after the repaired face image is subjected to face recognition.

2. The method according to claim 1, characterized in that Statistics on the first identity recognition accuracy rate that meets the occlusion area type in the historical face recognition data, including: The face recognition history data has an occlusion area record type; The occluded face recognition historical data whose occluded area record type overlaps with the occluded area type is extracted from the face recognition historical data, and the proportion of data with accurate identity recognition is counted, which is set as the first identity recognition accuracy rate.

3. The method according to claim 1, characterized in that Also includes: The updating of the occlusion area type is obtained by randomly deleting k types of the occlusion area type, wherein the number of occlusion area types is ≥ k ≥ 1, and the initial value of k is equal to 1; When the second identity recognition accuracy is less than the accuracy threshold, the occlusion area type is updated according to the k value.

4. The method according to claim 3, characterized in that Updating the occlusion region type according to the k value includes: According to the k value, the types of the occluded areas are combined and enumerated to obtain a plurality of type combinations; When at least one of the plurality of type combinations has not been randomly deleted, updating the updated occluded area type according to the type combination that has not been randomly deleted; When the plurality of type combinations have all been randomly deleted, the k value is increased by one and the occluded area type is updated.

5. The method according to claim 1, characterized in that When the face image is blocked, the method includes: Get the occlusion object type and occlusion constraint area type; Retrieving the occlusion object type and the occlusion constraint area type through an encoder to perform occlusion encoding on a preset face image, and obtaining an occlusion image set and an occlusion area type identifier set; Using the occlusion region type identification set as supervision data and the occlusion image set as input data, training an occlusion classifier; Processing the face image through the occlusion classifier to obtain the occlusion area type; When the occlusion area type is empty, the face image has no occlusion; When the occlusion region type is not empty, the face image has occlusion.

6. The method according to claim 5, characterized in that Repairing the face image according to the deleted area type identifier to obtain the repaired face image includes: Taking the preset face image as supervision and taking the occlusion region type identification set and the occlusion image set as input, training a face image repairer; According to the face image restorer, the deletion region type identifier and the face image are processed to obtain the restored face image.

7. The method according to claim 1, characterized in that Perform face recognition, including: Obtain a set of verification face images corresponding to the entered identity database; Traversing the verification face image set, comparing it with the repaired face image, and obtaining a verification similarity set; The input identity of the verification face image with the maximum verification similarity is extracted from the input identity database and set as the identity information.

8. A self-service system for user biometric identification, characterized in that: Used to implement the method according to any one of claims 1 to 7, comprising: A face image acquisition module, used to obtain a face image; A first recognition accuracy analysis module, for counting the first identity recognition accuracy rate that satisfies the occlusion area type in the face recognition history data when the face image is occluded; An occlusion area type deletion module, configured to randomly delete the occlusion area type when the first identity recognition accuracy is less than or equal to an accuracy threshold, to obtain an updated occlusion area type, wherein the updated occlusion area type has a deleted area type identifier; A second recognition accuracy analysis module, used to count the second identity recognition accuracy that meets the update occlusion area type in the face recognition history data; The biometric feature recognition module is used to repair the face image according to the deleted area type identifier when the second identity recognition accuracy is greater than the accuracy threshold, obtain the repaired face image, perform face recognition, and obtain identity information.

9. A self-service terminal for user biometric identification, characterized in that: Used to implement the method according to any one of claims 1 to 7.

Images