WO2022083653A1 - Method and apparatus for updating biometric library, and electronic device - Google Patents

Abstract

Provided in the present application are a method and apparatus for updating a biometric library, and an electronic device. The biometric library comprises at least one base biometric, and the method for updating the biometric library comprises: acquiring the similarity between an input biometric and a target base biometric, wherein the input biometric is the biometric extracted from multimedia data, and the target base biometric is the base biometric among the at least one base biometric that best matches the input biometric feature in terms of identity; when the similarity is greater than a first threshold, performing feature fusion on the input biometric and the target base biometric to obtain a fused feature; and updating the target base biometric as the fused feature.

Description

Method, device and electronic device for updating biometric database

This application is required to be submitted to the China Patent Office on October 20, 2020, the application number is 202011124636.9, and the title of the invention is "Method, Apparatus and Electronic Device for Updating Biometric Database", and the application should be submitted to the China Patent Office on January 15, 2021. The priority of the Chinese patent application No. 202110057029.3 and the title of the invention is "Method, Apparatus and Electronic Device for Updating Biometric Database", the entire contents of which are incorporated herein by reference.

technical field

The present application relates to the technical field of artificial intelligence, and in particular, to a method, an apparatus, an electronic device, and a computer-readable storage medium for updating a biometric database.

Background technique

In today's information age, how to accurately identify the individual's identity to protect the security of information has become a key social problem that must be solved. To identify an individual, biometric-based identification techniques can be employed. Biometric-based identification technology (referred to as biometric identification) refers to a technology that uses a computer to identify an individual by using the inherent physiological or behavioral characteristics of an organism. Among biometric-based identification technologies, face recognition technology is widely used due to its convenience and speed.

In the face recognition technology, the features of the face to be recognized are generally compared with the reference face features, and the identification result is obtained according to the similarity. However, face features may change greatly with changes in age, hairstyle, weather, season, etc., which may cause the features of the face to be recognized to be less similar to the reference face features due to the changes, and further This results in lower accuracy of face recognition. Therefore, how to improve the accuracy of face recognition is a technical problem that needs to be solved urgently. That is, how to improve the accuracy of individual identification using biometrics is a technical problem that needs to be solved urgently.

SUMMARY OF THE INVENTION

Embodiments of the present application provide a method, apparatus, electronic device, and computer-readable storage medium for updating a biometric database, so as to improve the accuracy of individual identification using biometrics.

According to an aspect of the embodiments of the present application, there is provided a method for updating a biometric database, the biometric database includes at least one reference biometric, and the method includes:

Obtaining the similarity between the input biometric feature and the target reference biometric feature, wherein the input biometric feature is a biometric feature extracted from multimedia data, and the target reference biometric feature is the at least one reference biometric feature that is the same as the The reference biometrics with the highest matching degree of identity of the input biometrics;

When the similarity is greater than the first threshold, feature fusion is performed on the input biometric feature and the target reference biometric feature to obtain a fused feature;

Updating the target reference biometrics to the fused features.

In some embodiments, the input biometric feature is obtained after biometric feature extraction is performed on a single picture containing the biometric feature; or

The input biometric feature is obtained by performing biometric feature extraction on a plurality of pictures containing biometric features and then performing biometric feature fusion.

In some embodiments, the updating of the target reference biometrics to the fused features includes:

Searching for the target reference biometric feature from the biometric database according to the identity of the target reference biometric feature;

Replace the target fiducial biometric with the fused feature.

In some embodiments, the obtaining the similarity between the input biometrics and the target reference biometrics includes: after completing the identification based on the input biometrics, obtaining the similarity between the input biometrics and the target reference biometrics Spend;

The first threshold value is greater than the second threshold value, and the second threshold value is: when performing identification based on the input biometric characteristics, the minimum similarity between the input biometric characteristic and the reference biometric characteristic indicating that the identification is successful.

In some embodiments, the method for updating the biometric database further includes: before acquiring the similarity between the input biometric and the target reference biometric, according to the similarity between the input biometric and the reference biometric degree, and the result information of whether the associated attributes of the input biometric feature and the reference biometric feature are consistent, and determine the target reference biometric feature;

Wherein, the confidence level of the similarity between the input biometric feature and the reference biometric feature is not less than a confidence level threshold, and the associated attribute includes at least one of gender, age, and physical shape category.

In some embodiments, the feature fusion of the input biometric feature and the target reference biometric feature to obtain the fused feature includes:

According to the function feat_out=feat_cap*(1-momentum)+feat_old*momentum, feature fusion is performed on the input biometrics and the target reference biometrics;

The feat_out is the feature after fusion, the feat_cap is the input biometric, the feat_old is the target reference biometric, the momentum is the momentum coefficient, and 0≤momentum<1.

In some embodiments, 0.9≤momentum<1.

In some embodiments, the biometric library includes a face feature library, a palm shape feature library, a skin feature library, a pinna feature library, a gait feature library, or a voice feature library.

According to another aspect of the embodiments of the present application, there is provided an apparatus for updating a biometric database, wherein the biometric database includes at least one reference biometric feature, and the apparatus includes:

an obtaining unit, configured to obtain the similarity between the input biometric feature and the target reference biometric feature, wherein the input biometric feature is a biometric feature extracted from multimedia data, and the target reference biometric feature is the at least one reference biological feature The reference biometric feature that matches the identity of the input biometric feature with the greatest degree of identity;

a fusion unit, configured to perform feature fusion on the input biological feature and the target reference biological feature when the similarity is greater than a first threshold to obtain a post-fusion feature;

An update unit, configured to update the target reference biological feature to the fusion feature.

According to yet another aspect of the embodiments of the present application, there is provided an electronic device, comprising: a memory and a processor coupled to the memory, the processor configured to execute any of the foregoing based on instructions stored in the memory The method for updating the biometric database described in the technical solution.

According to yet another aspect of the embodiments of the present application, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the method for updating a biometric database described in any of the foregoing technical solutions is implemented .

The above-mentioned embodiments of the present application can implement automatic updating of the biometric database according to program settings. Compared with the fixed reference biological feature in the related art, the reference biological feature of the embodiment of the present application can make dynamic adjustment following the change of the individual biological feature. The updated biometric database is applied to biometric identification, which can improve the accuracy of identification, thereby improving the security of information.

Of course, implementing the product or method of any embodiment of the present application does not necessarily need to achieve all the above advantages simultaneously.

Description of drawings

In order to more clearly illustrate the embodiments of the present application and the technical solutions of the prior art, the following briefly introduces the drawings required in the embodiments and the prior art. Obviously, the drawings in the following description are only the For some embodiments of the application, for those of ordinary skill in the art, other embodiments can also be obtained according to these drawings without creative efforts.

Fig. 1 is the flow chart of a kind of face recognition method in the related art;

2 is a flowchart of a method for updating a biometric database according to some embodiments of the present application;

3 is a flowchart of face recognition and update of face feature database in some embodiments of the present application;

4 is a block diagram of an apparatus for updating a biometric database according to some embodiments of the present application;

FIG. 5 is a block diagram of an electronic device according to some embodiments of the present application.

Detailed ways

In order to make the objectives, technical solutions, and advantages of the present application more clearly understood, the present application will be described in further detail below with reference to the accompanying drawings and examples. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present application.

Face recognition is a biometric recognition technology based on human facial features. The pre-established reference face features in the feature library are compared, and the recognition result of the face identity information is obtained according to the similarity.

As shown in FIG. 1, it is a face recognition method in the related art, and the method includes the following steps S01-S05.

In step S01, a picture including a face region collected by an image collection device is acquired.

In step S02, a face feature extraction algorithm is used to extract a face feature from a picture containing a face region, as an input face feature.

In step S03, the reference face features in the face feature database are traversed, and the similarity between the input face feature and each reference face feature is compared to obtain a similarity list.

In step S04, the similarity list is screened, and the similarity comparison data whose confidence level is not less than the confidence level threshold is retained as the screening result.

The above confidence level may reflect: the confidence of the result of the similarity comparison between the input face feature and the reference face feature.

In step S05, output the recognition result according to the reference face feature corresponding to the maximum similarity in the screening result. For example, the identification result may include the maximum similarity, the face image restored based on the reference face feature, the identification number of the individual corresponding to the reference face feature, and the like.

In the process of implementing the embodiments of the present application, the inventors of the present application found that since the reference facial features in the facial feature database are fixed, when the facial features change greatly with changes in age, hairstyle, weather, season, etc. The accuracy of face recognition results will be greatly reduced, which will bring information security risks.

In order to improve the accuracy of individual identification using biometrics, the embodiments of the present application provide a method, an apparatus, an electronic device, and a computer-readable storage medium for updating a biometrics database.

The technical solutions provided in the embodiments of this application can be applied to various scenarios that require biometric identification, such as security inspection monitoring scenarios, public place monitoring scenarios, access control monitoring scenarios, animal husbandry monitoring scenarios, etc. The applicable fields include but are not limited to Security, finance, border, customs, insurance and civil entertainment and other fields. Depending on the specific application scenarios of biometric identification, the biometric database may be, for example, a face feature library, a palm shape feature library, a skin feature library, a pinna feature library, a gait feature library, or a voice feature library, and so on. The biometric database may include only one reference biological feature, or may include multiple reference biological features, and the number of reference biological features in the biometric database may be increased or deleted according to user requirements.

The above-mentioned biometric database can be understood as a database for storing reference biometric characteristics. The above-mentioned reference biometrics can be understood as: pre-obtained biometrics used for feature comparison.

As shown in FIG. 2 , some embodiments of the present application provide a method for updating a biometric library, wherein the biometric library includes at least one reference biometric, and the method for updating the biometric library includes the following steps S1 to S3.

In step S1, the similarity between the input biometric feature and the target reference biometric feature is obtained, wherein the input biometric feature is a biometric feature extracted from multimedia data, and the target reference biometric feature is the at least one reference biometric feature and the input biometric feature. The benchmark biometrics with the greatest identity match.

The above-mentioned input biometrics can be understood as: the biometrics of the individual whose identity needs to be matched may be the biometrics extracted from the multimedia data including the above-mentioned individual.

The above identity matching degree can be understood as a parameter used to measure whether the identity of the individual reflected by the reference biometrics is consistent with the identity of the individual reflected by the input biometrics.

The above-mentioned target reference biometrics can be understood as: the characteristics determined from the reference biometrics, and the reflected individual matches the identity of the individual reflected by the above input biometrics. That is, the identity of the individual reflected by the target reference biometrics matches the identity of the individual reflected by the above input biometrics.

Specifically, from the reference biometric features stored in the biometric database, it is possible to determine the features of the reflected individual that match the identity of the individual reflected by the above input biometric features, as the target reference biological feature, and then obtain the target reference biological feature Similarity to the above input biometrics.

In some embodiments of the present application, when calculating the similarity between the input biometrics and the reference biometrics, the Euclidean distance, cosine similarity, Manhattan distance, etc. between the input biometrics and the reference biometrics may be calculated, as the above similarity between features.

In addition, the input biometrics and the reference biometrics can also be input into a pre-trained similarity calculation model to obtain the similarity between the input biometrics and the reference biometrics output by the model.

In some embodiments of the present application, the input biometric feature is obtained after biometric feature extraction is performed on a single picture containing the biometric feature. For example, the input biometric feature is extracted from a single photo taken by the camera that contains the biometric feature, or from a frame in the video stream captured by the camera that contains the biometric feature.

In other embodiments of the present application, the input biometric feature is obtained by performing biometric feature extraction on a plurality of pictures containing biometric features and then performing biometric feature fusion.

Specifically, a plurality of pictures containing biological features can be obtained, and biological features are extracted from each picture to obtain a plurality of biological features, and then feature fusion is performed on the above-mentioned plurality of biological features, and the fusion result is used as the input biological features.

For example, biometric feature extraction is performed on multiple frames captured by a camera that contain biometric features, and then these biometric features are fused to obtain input biometric features. In this way, when a certain picture or some pictures are blurry, the resolution is low or the shooting angle is unreasonable, the biometric fusion calculation can make the impact on the accuracy of the recognition result small or even negligible, thereby improving the accuracy of biometric recognition. accuracy.

In the embodiment of the present application, the reference biological feature of the biological feature database is an offline reference template obtained based on a deep learning algorithm. Deep learning is a type of machine learning, and machine learning is the only way to realize artificial intelligence. The concept of deep learning originated from the study of artificial neural networks, and a multilayer perceptron containing multiple hidden layers is a deep learning structure. Deep learning realizes distributed feature representation of data by combining low-level features to form more abstract high-level features that represent attribute categories. The motivation for studying deep learning is to build a neural network that simulates the human brain for analysis and learning, which imitates the mechanism of the human brain to parse data, such as images, sounds, and texts. Typical deep learning models include convolutional neural network models, deep neural network models, and stacked self-encoding network models.

In this embodiment of the present application, the target reference biometrics refers to: comparing each reference biometric feature in the biometric database with the input biometric features, wherein the reference biometric feature with the largest identity matching with the input biometric feature. Among them, the consideration factor of the identity matching degree includes the similarity of the biometric comparison.

For example, target baseline biometrics can be determined by:

Calculate the similarity between the input biometrics and each benchmark biometric in the biometric database, use the above similarity as the identity matching degree, and select the benchmark biometric with the highest similarity from the benchmark biometrics whose similarity exceeds the preset threshold as the identity matching degree. The target reference biometric whose reflected individual matches the identity of the individual as reflected by the above input biometrics.

In addition, the above-mentioned consideration factors of the identity matching degree may also include the consistency of other attributes associated with the user identity, for example, the consistency of gender, age group, physical form category, and the like. In the biometric database, the attributes associated with the user identity (hereinafter referred to as the associated attributes) of the reference biometrics can be pre-calculated by applying a certain algorithm and stored in the memory. In biometric identification, the associated attributes of the input biometrics can be calculated by using a certain algorithm. That is, the target reference biometrics can be understood as: in the biometric database, the reference biometrics with the highest degree of identity matching with the input biometrics and consistent associated attributes.

In some embodiments of the present application, the method for updating the biometric database further includes: before acquiring the similarity between the input biometric and the target reference biometric, according to the similarity between the input biometric and the reference biometric, and The result information of whether the associated attributes of the input biometrics and the reference biometrics are consistent, and determine the target reference biometrics; wherein, the confidence level of the similarity between the input biometrics and the reference biometrics is not less than the confidence level threshold, and the associated attributes include gender, At least one of age group and physical form category.

Wherein, the above confidence level may reflect: the confidence of the similarity between the acquired input biometrics and the reference biometrics.

Specifically, the similarity between the input biometrics and each reference biometric in the biometric database can be obtained, and the confidence level of the similarity is determined to be greater than or equal to a preset confidence level threshold, the similarity is the largest, and the associated attribute is the same as the input biometrics. Consistent benchmark biometrics as the target benchmark biometrics with the greatest identity match.

In some embodiments of the present application, the similarity between the input biometrics and each reference biometric in the biometric database can be obtained, and the reference biometrics whose confidence level of the similarity is greater than or equal to a preset confidence level threshold are selected from the comparison, and from From the selected benchmark biometrics, select the benchmark biometrics whose associated attributes are consistent with the input biometrics, and determine the feature with the greatest similarity with the input biometrics from the selected benchmark biometrics, as the target benchmark biometric with the greatest identity matching. feature.

In addition, it is also possible to search for the reference biometrics whose associated attributes are consistent with the input biometrics from the biometrics database, and use them as candidate reference biometrics to obtain the similarity between each candidate reference biometric and the input biometrics, and then filter similarities. The confidence level of the degree is greater than or equal to the preset confidence level threshold, and the feature with the greatest similarity with the input biometrics is determined from the filtered benchmark biometrics as the target reference biometric with the greatest identity matching degree.

In some embodiments of the present application, the above step S1 is performed after completing the identification based on the input biometrics. In other embodiments of the present application, the above-mentioned step S1 may also be performed in the process of performing identification based on the input biometrics, or before performing the identification based on the inputted biometrics.

After the identification of the input biometrics is completed, the identification result is generally output. The identification result may include the identity verification result of the input biometrics, and in addition, may also include the similarity of the input biometrics compared with the target reference biometrics, the identity of the individual corresponding to the target reference biometrics, based on The image of the individual restored by the target reference biometrics, and other associated attributes, such as gender, age, physical shape category, etc.

In some embodiments of the present application, after each identification based on the input biometrics is completed, the similarity between the input biometrics and the target reference biometrics is obtained. That is, every time the biometric identification program is executed, after the identification is completed, a program for updating the biometric database is started.

In other embodiments of the present application, each time a set period of time passes, the identification result outputted by the identification based on the input biometric feature within the set period of time may be obtained, and the identification result includes the input biometric feature and the input biometric feature. Similarity of target baseline biometric alignments. That is, the procedure for updating the biometric database is started at a predetermined frequency. The set time period can be determined in conjunction with system processing performance and update requirements of the biometric database.

Returning to FIG. 2, in step S2, when the similarity between the input biometrics and the target reference biometrics is greater than the first threshold, feature fusion is performed on the input biometrics and the target reference biometrics to obtain fused features.

In some embodiments of the present application, the first threshold is preset and greater than the second threshold, and the second threshold is the minimum similarity between the input biometrics and the reference biometrics indicating successful identification when performing identification based on the input biometrics Spend. Application scenario settings of some embodiments of the present application: when the similarity between the input biometrics and the target reference biometrics is not less than the second threshold, the input biometric identity verification is allowed to pass; otherwise, the input biometric identity is not allowed Verification passed.

For example, in one embodiment, the second threshold is 95% and the first threshold is 97%. When the similarity between the input biometric and the target reference biometric is not less than 95%, the input biometric authentication is allowed to pass, otherwise, the input biometric authentication is not allowed to pass. When the similarity based on the comparison between the input biometrics and the target reference biometrics is greater than 97%, the biometric database is updated; otherwise, the biometric database is not updated. That is, compared with the judgment of the input biometric identity, a higher similarity requirement is put forward for the update of the biometric database, so that the accuracy of the update of the biometric database can be improved.

In deep learning, both input biometrics and reference biometrics are vectors, and feature fusion refers to a method of transforming two or more biometric vectors into one biometric vector.

In some embodiments of the present application, feature fusion is performed on the input biometrics and the target reference biometrics to obtain the fused features, including:

According to the function feat_out=feat_cap*(1–momentum)+feat_old*momentum, the feature fusion is performed on the input biometrics and the target benchmark biometrics; among them, feat_out is the fused feature, feat_cap is the input biometric, feat_old is the target benchmark biometric, momentum is the momentum coefficient, and 0≤momentum<1.

In some embodiments, the momentum coefficient is greater than or equal to 0.9 and less than 1. For example, in one embodiment, the momentum value is 0.95. In this way, the update of the target reference biometrics is gentle and subtle, resulting in higher accuracy of biometric identification.

In other embodiments of the present application, the feature fusion of the input biometrics and the target reference biometrics to obtain the fused features includes: splicing the vector components of the input biometrics and the target reference biological features to obtain the fused features.

Returning to Fig. 2, in step S3, the target reference biometric feature is updated to the fused feature.

In the biometric database, each reference biometric corresponds to an identity document (ID). In some embodiments, the above step S3 includes:

Find the target reference biometrics from the biometric database according to the identification of the target reference biometrics;

Replace the target baseline biometrics with the fused features.

As shown in FIG. 3 , the methods for updating a biometric database according to some embodiments of the present application are used in a face recognition scenario, and the method process for face recognition and updating the facial feature database includes the following steps S21 to S33 .

In step S21, a plurality of pictures including a face region captured by the image acquisition device are acquired.

For example, before entering the office area of the company, user Wang needs to complete the punch-in through an attendance punch-in machine with face recognition function. The camera of the attendance punching machine captured a series of video frames P1, P2, P3..., P12 including the face area of Wang.

In step S22, according to the set scoring rules, quality scores are performed on the plurality of pictures including the face region obtained in the above step S21, and pictures whose quality scores are lower than the set threshold are filtered out.

For example, the video frame P11 does not pass the quality score because the face area is blurred, and the video frame P12 fails the quality score because the face area is incomplete, then the video frames P11 and P12 are filtered out in this step S22.

In step S23, for each picture including a face region, a face feature extraction algorithm is used to extract the face feature from the picture. For example, face features F1, F2, F3..., F10 are extracted from video frames P1, P2, P3..., P10, respectively.

In step S23, the facial feature extraction algorithm used is not limited.

For example, in some embodiments, the facial feature extraction algorithm adopts an Active Shape Model (ASM) algorithm. ASM is an algorithm based on Point Distribution Model (PDM). In the PDM algorithm, the geometric shapes of objects with similar shapes, such as faces, hands, hearts, lungs, etc., can be represented by a shape vector formed by concatenating the coordinates of several key feature points in sequence. The process of using the ASM algorithm to extract face features from pictures includes: using the classifier to detect the face area; using the trained model to find a fixed number of feature points (for example, 68 feature points) on the face; recording the coordinates of each feature point position, and in turn concatenate to form a shape vector.

For example, in other embodiments, the facial feature extraction algorithm adopts an Active Appearance Model (Active Appearance Model, AAM) algorithm. On the basis of ASM algorithm, AAM algorithm further performs statistical modeling on texture, and further fuses the two statistical models of shape and texture into an appearance model. After the AAM algorithm unifies the dimensions of the shape and texture features, the modeling and search process is basically the same as that of the ASM.

In step S24, the face features F1, F2, F3..., F10 extracted in step S23 are subjected to feature fusion to obtain the input face feature F.

For example, in the aforementioned series of video frames P1, P2, P3..., P10, the picture of the video frame P2 is slightly blurred. If the face feature F2 is simply extracted from the video frame P2 as the input face feature for face recognition, the accuracy of the recognition result may be affected to a certain extent. In this embodiment, feature fusion processing is performed on the facial features of the series of video frames, and the influence of the image quality of the video frame P2 on the input facial features can be ignored, so the accuracy of face recognition can be improved.

In step S25, traverse the reference face features M1, M2, M3... in the face feature database, compare the similarity between the input face feature F and each reference face feature, and obtain a similarity list as shown in Table 1 below shown.

Table 1 Similarity List

In step S26, the similarity list is screened, and the similarity comparison data whose confidence level is not less than the confidence level threshold is retained as the first screening result, as shown in Table 2 below.

Table 2 List of first screening results

Benchmark facial features M1 M6 M8 Compare similarity 98% 99% 93%

When sampling to estimate population parameters, the conclusion is uncertain due to the randomness of the sample. The confidence level is the probability that the population parameter value falls within a certain interval of the sample statistic value. In the similarity list, if the confidence level of the similarity comparison data is less than the confidence level threshold, the face recognition result is considered unreliable, and the comparison data is eliminated; if the confidence level of the similarity comparison data is not less than the confidence level threshold, The face recognition result is considered credible, and the comparison data is retained.

In step S27, the first screening result is screened again, and the comparison data of the reference face feature whose associated attribute is consistent with the input face feature F is retained as the second screening result, as shown in Table 3 below. In some embodiments, the associated attribute includes at least one of gender, age group, and physical form categories.

Table 3 The second filter result list

Benchmark facial features M1 M8 Compare similarity 98% 93%

For example, the gender corresponding to the reference face feature M6 in the above Table 2 is inconsistent with the gender obtained based on the input face feature, therefore, the comparison data related to the reference face feature M6 is screened out in this step S27 .

In step S28, the reference face feature corresponding to the maximum similarity in the second screening result is used as the target reference face feature M1, and a recognition result is output based on the target reference face feature M1. The recognition result includes the above-mentioned maximum similarity (98%), the face image restored based on the target benchmark face feature M1, and Wang's identity, such as Wang's job number 00345.

Since the reference face features whose confidence level is less than the confidence level threshold have been screened out in the aforementioned step S27, and the reference face features whose associated attributes are inconsistent with the input face feature F have been screened out in the aforementioned step S28, therefore, the step S28 The amount of calculation is greatly reduced, and the output results are more accurate.

In step S29, the similarity between the input biometrics and the target reference biometrics is obtained.

In step S30, it is judged whether the similarity (98%) between the input biometric feature and the target reference biometric feature is greater than the first threshold (eg, the first threshold is 95%), if so, the process goes to step S31, otherwise, the process ends.

In step S31, the target reference face feature M1 is found from the face feature database according to the identity identifier corresponding to the target reference face feature M1, such as Wang's job number 00345.

In step S32, feature fusion is performed on the input face feature F and the target reference face feature M1 to obtain a fused feature M1'.

In step S33, the target reference face feature M1 in the face feature library is updated to the fused feature M1', that is, the target reference face feature M1 in the face feature library is replaced with the fused feature M1'.

As shown in FIG. 4 , some embodiments of the present application further provide a device 400 for updating a biometric database, where the biometric database includes at least one reference biometric, and the device 400 for updating a biometric database includes:

The obtaining unit 41 is used to obtain the similarity between the input biological feature and the target reference biological feature, wherein the input biological feature is the biological feature extracted from the multimedia data, and the target reference biological feature is at least one reference biological feature and the input biological feature. Benchmark biometrics with the highest matching degree of identity of the features;

The fusion unit 42 is configured to perform feature fusion on the input biological feature and the target reference biological feature when the similarity is greater than the first threshold to obtain the fused feature;

The updating unit 43 is configured to update the target reference biological feature to the fused feature.

The above-mentioned embodiments of the present application can implement automatic updating of the biometric database according to program settings. Compared with the fixed reference biological feature in the related art, the reference biological feature of the embodiment of the present application can make dynamic adjustment following the change of the individual biological feature. The updated biometric database is applied to biometric identification, which can improve the accuracy of identification, thereby improving the security of information.

As shown in FIG. 5 , some embodiments of the present application further provide an electronic device 500 , comprising: a memory 51 and a processor 52 coupled to the memory 51 , the processor 52 is configured to be based on instructions stored in the memory 51 , The method for updating a biometric database as in any of the preceding embodiments is performed.

The memory 51 may include random access memory (Random Access Memory, RAM), or may include non-volatile memory (Non-Volatile Memory, NVM), such as at least one disk storage. The memory 51 may also be at least one storage device located remotely from the aforementioned processor 52 .

The above-mentioned processor 52 can be a general-purpose processor, including a central processing unit (Central Processing Unit, CPU), a network processor (Network Processor, NP), etc.; it can also be a digital signal processor (Digital Signal Processor, DSP), dedicated Integrated circuit (Application Specific Integrated Circuit, ASIC), Field-Programmable Gate Array (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

Some embodiments of the present application further provide a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, implements the method for updating a biometric database according to any of the foregoing technical solutions.

In addition, in still other embodiments of the present application, a computer program product including instructions is also provided, when it runs on a computer, the computer causes the computer to execute the method for updating a biometric database in any of the foregoing embodiments.

In the above-mentioned embodiments, it may be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented in software, it can be implemented in whole or in part in the form of a computer program product. A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the procedures or functions according to the embodiments of the present application are generated in whole or in part. The computer may be a general purpose computer, a special purpose computer, a computer network, or other programmable device. Computer instructions may be stored in or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions may be transmitted from a website site, computer, server, or data center over a wire (e.g. coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (eg, infrared, wireless, microwave, etc.) to another website site, computer, server, or data center. A computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server, a data center, or the like that includes an integration of one or more available media. Useful media may be magnetic media (eg, floppy disks, hard disks, magnetic tapes), optical media (eg, DVD), or semiconductor media (eg, Solid State Disk (SSD)), among others.

It should be noted that, in this document, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any relationship between these entities or operations. any such actual relationship or sequence exists. Moreover, the terms "comprising", "comprising" or any other variation thereof are intended to encompass a non-exclusive inclusion such that a process, method, article or device that includes a list of elements includes not only those elements, but also includes not explicitly listed or other elements inherent to such a process, method, article or apparatus. Without further limitation, an element qualified by the phrase "comprising a..." does not preclude the presence of additional identical elements in the process, method, article, or device that includes the element.

Each embodiment in this specification is described in a related manner, and the same and similar parts between the various embodiments may be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the apparatus, electronic device, and computer-readable storage medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and reference may be made to some descriptions of the method embodiments for related parts.

The above are only preferred embodiments of the present application, and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application shall be included in the protection of the present application. within the range.

Claims (11)

A method for updating a biometric database, the biometric database comprising at least one reference biometric, characterized in that the method comprises: Obtaining the similarity between the input biometric feature and the target reference biometric feature, wherein the input biometric feature is a biometric feature extracted from multimedia data, and the target reference biometric feature is the at least one reference biometric feature that is the same as the The reference biometrics with the highest matching degree of identity of the input biometrics; When the similarity is greater than the first threshold, feature fusion is performed on the input biometric feature and the target reference biometric feature to obtain a fused feature; Updating the target reference biometrics to the fused features. The method for updating a biometric database according to claim 1, wherein, The input biometric feature is obtained after biometric feature extraction is performed on a single picture containing the biometric feature; or The input biometric feature is obtained by performing biometric feature extraction on a plurality of pictures containing biometric features and then performing biometric feature fusion. The method for updating a biometric database according to claim 1, wherein the updating the target reference biometric feature to the post-fusion feature comprises: Searching for the target reference biometric feature from the biometric database according to the identity of the target reference biometric feature; Replace the target fiducial biometric with the fused feature. The method for updating a biometric database according to claim 1, wherein, The obtaining the similarity between the input biometrics and the target reference biometrics includes: after completing the identity recognition based on the input biometrics, acquiring the similarity between the input biometrics and the target reference biometrics; The first threshold value is greater than the second threshold value, and the second threshold value is: when performing identification based on the input biometric characteristics, the minimum similarity between the input biometric characteristic and the reference biometric characteristic indicating that the identification is successful. The method for updating a biometric database according to claim 1, further comprising: Before acquiring the similarity between the input biometrics and the target reference biometrics, according to the similarity between the input biometrics and the reference biometrics, and the similarity between the input biometrics and the reference biometrics The result information of whether the associated attributes are consistent to determine the target benchmark biometrics; Wherein, the confidence level of the similarity between the input biometric feature and the reference biometric feature is not less than a confidence level threshold, and the associated attribute includes at least one of gender, age, and physical shape category. The method for updating a biometric database according to claim 1, wherein the feature fusion of the input biometrics and the target reference biometrics to obtain the fused features, comprising: According to the function feat_out=feat_cap*(1-momentum)+feat_old*momentum, feature fusion is performed on the input biometrics and the target reference biometrics; The feat_out is the feature after fusion, the feat_cap is the input biometric, the feat_old is the target reference biometric, the momentum is the momentum coefficient, and 0≤momentum<1. The method for updating a biometric database according to claim 6, wherein 0.9≤momentum<1. The method for updating a biometric database according to any one of claims 1-7, wherein the biometric database includes a face feature database, a palm shape feature database, a skin feature database, a pinna feature database, and a gait feature database. Signature library or sound signature library. A device for updating a biometric database, the biometric database comprising at least one reference biometric, characterized in that the device includes: an obtaining unit, configured to obtain the similarity between the input biometric feature and the target reference biometric feature, wherein the input biometric feature is a biometric feature extracted from multimedia data, and the target reference biometric feature is the at least one reference biological feature The reference biometric feature that matches the identity of the input biometric feature with the greatest degree of identity; a fusion unit, configured to perform feature fusion on the input biological feature and the target reference biological feature when the similarity is greater than a first threshold to obtain a post-fusion feature; An update unit, configured to update the target reference biological feature to the fusion feature. An electronic device, comprising: a memory and a processor coupled to the memory, the processor configured to execute any one of claims 1-8 based on instructions stored in the memory The described method for updating a biometric database. A computer-readable storage medium, characterized in that a computer program is stored thereon, and when the computer program is executed by a processor, the method for updating a biometric database according to any one of claims 1-8 is implemented.

Images