WO2023036322A1

WO2023036322A1 - Rapid authentication method based on finger vein biometric recognition technology, and device and medium

Info

Publication number: WO2023036322A1
Application number: PCT/CN2022/118264
Authority: WO
Inventors: 汤浩新; 彭冠莲
Original assignee: 广州广电运通金融电子股份有限公司
Priority date: 2021-09-13
Filing date: 2022-09-09
Publication date: 2023-03-16
Also published as: CN113792649A

Abstract

The present disclosure relates to a rapid authentication method based on finger vein biometric recognition technology, and a device and a medium. The rapid authentication method comprises: distributing finger vein templates in a registration template library to a front computing node or a plurality of parallel computing nodes according to a preset distribution rule; and in response to a recognition request that is sent by a vein terminal, sending a collected vein image to the corresponding front computing node or parallel computing nodes to perform feature matching processing, so as to output a final authentication result. In the present disclosure, the biometric recognition efficiency is improved by means of a combination of parallel computation, front recognition and user mode self-learning.

Description

A fast authentication method, device and medium based on finger vein biometric technology

This disclosure claims the priority of the Chinese patent application with the application number 202111069323.2 and the title of the invention "a fast authentication method, device and medium based on finger vein biometric technology" submitted to the China Patent Office on September 13, 2021. The entire contents are incorporated by reference in this disclosure.

technical field

The present disclosure relates to the technical field of finger vein recognition, and in particular to a fast authentication method, device and storage medium based on finger vein biometric technology.

Background technique

At present, how to avoid identity forgery and quickly and accurately provide identity authentication services is an urgent problem in today's increasingly demanding identity authentication. The continuous development of biometric technology has provided strong support and guarantee for the application of identity verification. Among them, finger vein recognition has the characteristics of high anti-counterfeiting, high accuracy, live detection, strong adaptability and ease of use. Finger vein recognition based on finger vein recognition technology Combined with big data analysis applications, the vein recognition platform provides more diverse, accurate and convenient services.

Finger vein authentication technology is a new biometric identification technology, which is developed based on the principle that blood flowing in human fingers can absorb light of a specific wavelength. Using specific light irradiation, the characteristic image of the finger vein can be obtained, and the biological characteristics of the finger vein can be obtained by further analyzing and processing the image, and the identification and authentication can be realized by comparing it with the pre-registered characteristics.

However, the existing vein comparison needs to take vein images in real time, extract eigenvalues, use advanced filtering, image binarization, and thinning methods to extract features from digital images, compare them with the eigenvalues of finger vein templates, and use complex matching Algorithms match finger vein features to identify individuals and confirm their identity. However, the process requires a large amount of calculation. When the registration template library to be matched is larger and the number of templates is larger, the time required to complete a recognition is longer, resulting in the speed of the recognition and authentication process being unable to increase.

Contents of the invention

(1) Technical problems to be solved

The technical problem to be solved in the present disclosure is to solve the problem that the speed of the existing identification and authentication process cannot be improved.

(2) Technical solution

In order to solve the above technical problems, an embodiment of the present disclosure provides a rapid authentication method based on finger vein biometric technology, including:

Distribute the finger vein templates in the registered template library to a pre-computing node or multiple parallel computing nodes according to the preset distribution rules;

In response to the identification request sent by the vein terminal, the acquired vein image is sent to the corresponding pre-computing node or parallel computing node for feature matching processing to output the final authentication result.

Further, the distribution rule corresponding to the parallel computing nodes is to randomly distribute all finger vein templates in the registered template library to multiple parallel computing nodes.

Further, the method of sending the vein image to the parallel computing node for feature matching processing to output the final authentication result is as follows:

After receiving the recognition request, send the recognition request to all parallel computing nodes for parallel processing by broadcasting, so that all parallel computing nodes will compare the vein images and the finger vein templates stored in their respective nodes, and use The matching result returned fastest is output as the final authentication result.

Further, the distribution rule corresponding to the pre-computing node is to distribute the finger vein templates in the registration template library whose frequency of use is greater than a preset value to a pre-computing node.

Further, the method of sending the vein image to the pre-calculation node for feature matching processing to output the final authentication result is as follows:

After receiving the identification request, the identification request is sent to the front-end computing node for priority identification processing, if the real-time collected vein image cannot obtain a matching result with any finger vein template in the front-end computing node, Then match the vein image with the remaining finger vein templates in the registration template library to output the final authentication result.

Further, when adding the finger vein template to the front computing node, it also includes:

Judging the number of finger vein templates in the front computing node, if the number of finger vein templates in the front computing node is within the preset number, adding the finger vein template to the front computing node If the number of finger vein templates in the preceding computing node reaches the preset number, the finger vein template is added to the preceding computing node and the last matching time in the preceding computing node is The farthest finger vein template is moved out of the preceding computing node.

Further, it also includes:

The real-time collected vein image that has passed the matching authentication and the image quality reaches the preset score is stored in the front-end computing node as the finger vein template of the recognition object, and is preferentially compared with the vein image of the recognition object next time. Feature matching is performed on the finger vein template in the pre-computing node to obtain the final authentication result.

Further, when storing the vein image of any identified object collected in real time in the front computing node, it also includes:

Judging whether the number of finger vein templates of the identified object in the front-end computing node reaches a threshold, and if it exceeds the threshold, only keep a specified number of finger vein templates with the highest image quality score and the latest acquisition time.

In a second aspect, an embodiment of the present disclosure provides an electronic device, including: a processor, a memory, and a computer program stored on the memory and operable on the processor, and the processor executes the computer program Realize above-mentioned fast authentication method based on finger vein biometric technology at the same time.

In a third aspect, an embodiment of the present disclosure provides a storage medium on which a computer program is stored, and when the computer program is executed, the above-mentioned rapid authentication method based on finger vein biometric technology is implemented.

(3) Beneficial effects

Compared with the prior art, the above-mentioned technical solutions provided by the embodiments of the present disclosure have the following advantages: the present disclosure can pre-allocate the templates in the registered template library to different computing nodes, and use different computing nodes to realize parallel computing and pre-identification , User mode self-learning combined to reduce a large amount of invalid calculations and improve the efficiency of finger vein biometrics.

Description of drawings

FIG. 1 is a schematic diagram of the server architecture of the disclosed parallel computing solution;

Fig. 2 is a schematic diagram of the finger vein template distribution rule of the parallel computing scheme of the present disclosure;

3 is a schematic flow diagram of the disclosed parallel computing solution;

FIG. 4 is a schematic flow diagram of the pre-identification scheme of the present disclosure;

FIG. 5 is a schematic flowchart of the user mode self-learning solution of the present disclosure.

Detailed ways

Below, the present disclosure will be further described in conjunction with the accompanying drawings and specific implementation methods. It should be noted that, on the premise of not conflicting, the various embodiments described below or the technical features can be combined arbitrarily to form a new embodiment. .

This embodiment provides a fast authentication method based on finger vein biometric technology. This embodiment uses the combination of parallel computing, pre-identification, and user mode self-learning to reduce the processing time of finger vein recognition; The rapid authentication method of the example mainly includes the following steps:

Step S1: Distribute the finger vein templates in the registered template library to a pre-computing node or multiple parallel computing nodes according to the preset distribution rules;

Step S2: In response to the identification request sent by the vein terminal, the collected vein image is sent to the corresponding pre-computing node or parallel computing node for feature matching processing to output the final authentication result.

In this embodiment, the corresponding distribution rules and corresponding authentication methods can be configured according to the size of the registration template library, specifically: due to the limited computing power of a single node, it is necessary to form a computing cluster with multiple nodes to improve computing power, if the finger stored in the registration template library If the number of vein templates is relatively large and the memory of the registration template library is large, the parallel computing solution can be used to complete the fast authentication operation; if the registration template library is relatively small, the pre-identification or user mode self-learning solution can be used to complete the fast authentication operation .

As shown in Figure 1, Figure 2 and Figure 3, the parallel computing solution in this embodiment can achieve the purpose of fast authentication through the registration template server, parallel computing node and identification server, specifically, the registration template server can pre-register the entire All pre-stored finger vein templates in the registration template library are randomly distributed to different parallel computing nodes for storage; at the same time, the registration template server can receive adjustment instructions and increase the templates loaded in each parallel computing node according to the instructions Or delete, so as to achieve the effect of real-time dynamic adjustment of intelligent adaptation. In this embodiment, a plurality of parallel computing nodes can be set according to the actual situation. After receiving the recognition request, each parallel computing node will perform feature extraction on the acquired vein image, and compare the vein features with the parallel computing node. Compare and match the finger vein templates stored in the database, and return the calculation results in time; and multiple parallel computing nodes can perform parallel computing at the same time, thereby shortening the overall time of matching authentication and improving authentication efficiency. The recognition server in this embodiment is used to receive the recognition request sent by the finger vein terminal, and then send the recognition request and the collected vein image to the entire node cluster through broadcasting, so that all parallel computing nodes can perform parallel computing. When multiple parallel computing nodes get matching results, the fastest returned matching result will be output as the final authentication result.

Because in real usage scenarios, the finger vein templates corresponding to users with high usage rate only account for a small part of the entire template library. For example, a user registers 6 fingers (users usually register fingers as the index finger, middle finger, and ring finger of both hands, a total of 6 fingers), and in actual daily use, only one or two of them are often used. Therefore, most registration templates are non-hot data. If the traversal matching is performed each time the recognition is performed, the calculation amount of this part of the data is likely to be invalid calculation. Therefore, in order to improve authentication efficiency and reduce some invalid calculations, as shown in Figure 4, the hot template data set can be obtained through specific logic, and the data set can be identified as a front. If the matching is successful from the front template library, then The recognition process is completed. If there is no match in the front-end template library, the second round of recognition is performed with the complete registration template library, which can reduce a large amount of invalid calculations and greatly improve the recognition efficiency.

However, in this embodiment, the front-end computing node is used to replace the above-mentioned front-end template library to complete the matching operation. Specifically, the number of uses of each finger vein template in the registered template library is counted in advance, and the usage frequency is higher than the preset value. The finger vein template is marked as a hotspot template, and the hotspot template is distributed to the front-end computing node; after receiving the identification request sent by the finger vein terminal, the identification request is sent to the front-end computing node for priority identification processing. Configure the computing node to compare the features of the vein image collected by the finger vein terminal with the stored hotspot template. If any hotspot template matches the vein image, the match will be successful and the corresponding authentication result will be output; if the vein image collected in real time cannot If the matching result is obtained with any finger vein template in the pre-computing node, then the vein image is matched with other finger vein templates in the registration template library to output the final authentication result. After the hotspot template is added to the front computing node in the registration template library, it can be marked or deleted, so that when the vein image and the registration template library are compared and matched, repeated comparison of the hotspot template can be avoided, thereby further improving recognition efficiency.

If it cannot be matched through the front-end computing node/front-end template library, but can be matched through the complete registration template library, the successfully matched template in the registration template library will be automatically added to the front-end computing node/front-end template library as hot data . According to the actual test, the front-end computing node/front-end template library in a single thread can be set to 2000 templates, and the recognition time is within 1 second, which meets the performance and business requirements. Therefore, if the size of the current pre-computing node/pre-template library exceeds the set size, the template with the farthest last matching time in the pre-compute node/pre-template library will be removed. Specifically: before adding the finger vein template to the pre-computing node/pre-template library, it is necessary to judge the number of finger vein templates in the pre-computing node/pre-template library. If the number of finger vein templates in the node/pre-template library is within the preset number, finger vein templates can be added to the pre-computing node/pre-template library; if the pre-computing node/pre-template When the number of finger vein templates in the template library reaches the preset number, the finger vein templates are added to the pre-computing node/pre-template library and the last matching The finger vein template with the farthest time is removed from the front-end computing node/front-end template library.

In addition to the above two identification schemes, this embodiment can also learn the individual usage habits of each user through the self-learning scheme during the user's use process, and adapt the identification logic to each user's different usage habits to achieve system coordination. User habits, not user cooperation with system limitations.

Due to the difference between near-infrared light illumination and optical imaging sensors, finger vein images collected by different devices in different scenarios are different. In addition, different irradiation intensities such as near-infrared light will also cause nonlinear degradation to the finger vein image. Since nonlinear degradation cannot be directly modeled, the influence of light intensity is eliminated. In addition, different image sensors have different differences, which will also lead to different imaging results for the same finger under the same lighting. Therefore, in the actual application of finger veins, there are unavoidable differences in finger vein images collected by different devices. As shown in Figure 5, while providing recognition services in actual finger vein applications, samples with higher matching scores are added to the pre-template library/pre-template library as templates for the same recognition object to adapt to the current factors that affect recognition , so as to greatly improve the recognition rate.

Specifically, the quality score of the vein image that has passed the matching authentication is performed in advance. Generally, the vein image that has passed the matching authentication can be scored as a standard score. If the resolution of the vein image that has passed the matching authentication reaches the preset value, it can be evaluated. It is a score above the standard score, for example, the standard score is 75 points, if the quality score of a vein image that passes the matching authentication is rated as 85 points, then the vein image can be considered as a high-quality image and can be used as the basis for template training Data, that is, the vein image is added to the front template library/front computing node as the template of the object, and it will be preferentially included in the front template library/front computing node when the vein image of the identified object is collected next time Feature matching is performed on the finger vein template to improve the recognition success rate. According to actual data statistics, the recognition rate can be increased from 78% to 92% in this way.

In order to avoid the distribution balance of the front-end template library/front-end computing node, when storing the vein image of any recognized object collected in real time in the front-end template library/front-end computing node, it is also necessary to determine the Set whether the number of finger vein templates of the identified object in the calculation node reaches the threshold, if it exceeds the threshold, only keep the specified number of finger vein templates with the highest image quality score and the latest collection time; if there are more than 3 templates for the same identified object, Then only keep the three with the highest score and the latest time.

The three schemes of parallel computing, pre-identification, and user mode self-learning provided in this embodiment can operate independently or be used in combination with each other, which can significantly improve the efficiency of biometric identification.

Embodiment two

This embodiment provides an electronic device, which may be a device such as a venous terminal, a background server, etc., which includes a processor, a memory, and a computer program stored on the memory and operable on the processor. When the processor executes the computer program, the fast authentication method based on finger vein biometric technology in Embodiment 1 is implemented; in addition, this embodiment also provides a storage medium on which a computer program is stored, and the computer program is executed Realize above-mentioned fast authentication method based on finger vein biometric technology at the same time.

The equipment and storage medium in this embodiment are based on two aspects under the same inventive concept as the method in the foregoing embodiments. The implementation process of the method has been described in detail above, so those skilled in the art can understand clearly from the foregoing description The structure and implementation process of the equipment in this embodiment are well understood, and for the sake of brevity of the description, details are not repeated here.

The above-mentioned embodiments are only preferred embodiments of the present disclosure, and cannot be used to limit the protection scope of the present disclosure. Any insubstantial changes and substitutions made by those skilled in the art on the basis of the present disclosure belong to the scope of the present disclosure. Scope of protection claimed.

Industrial Applicability

The fast authentication method based on finger vein biometric technology provided by this disclosure can pre-allocate templates in the registration template library to different computing nodes, and use different computing nodes to realize parallel computing, pre-identification, and self-learning of user patterns. Combination, thereby reducing a large amount of invalid calculations, improving the efficiency of finger vein biometrics, and has strong industrial applicability.

Claims

A fast authentication method based on finger vein biometric technology, characterized in that it includes:

Distribute the finger vein templates in the registered template library to a pre-computing node or multiple parallel computing nodes according to the preset distribution rules;

In response to the identification request sent by the vein terminal, the acquired vein image is sent to the corresponding pre-computing node or parallel computing node for feature matching processing to output the final authentication result.
The fast authentication method based on finger vein biometrics according to claim 1, wherein the distribution rule corresponding to the parallel computing nodes is to randomly distribute all finger vein templates in the registered template library to multiple parallel computing nodes. in the compute node.
The fast authentication method based on finger vein biometrics according to claim 2, wherein the method of sending the vein image to a parallel computing node for feature matching processing to output the final authentication result is:

After receiving the recognition request, send the recognition request to all parallel computing nodes for parallel processing by broadcasting, so that all parallel computing nodes will compare the vein images and the finger vein templates stored in their respective nodes, and use The matching result returned fastest is output as the final authentication result.
The fast authentication method based on finger vein biometrics according to claim 1, wherein the distribution rule corresponding to the pre-computing node is to divide the finger vein templates in the registration template library whose usage frequency is greater than the preset value Centrally sent to a pre-computing node.
The fast authentication method based on finger vein biometrics according to claim 4, wherein the method of sending the vein image to the pre-computing node for feature matching processing to output the final authentication result is:

After receiving the identification request, the identification request is sent to the front-end computing node for priority identification processing, if the real-time collected vein image cannot obtain a matching result with any finger vein template in the front-end computing node, Then match the vein image with the remaining finger vein templates in the registration template library to output the final authentication result.
The fast authentication method based on finger vein biometrics according to claim 4, wherein adding the finger vein template to the front computing node also includes:

Judging the number of finger vein templates in the front computing node, if the number of finger vein templates in the front computing node is within the preset number, adding the finger vein template to the front computing node If the number of finger vein templates in the preceding computing node reaches the preset number, the finger vein template is added to the preceding computing node and the last matching time in the preceding computing node is The farthest finger vein template is moved out of the preceding computing node.
The fast authentication method based on finger vein biometrics according to claim 1, further comprising:

The real-time collected vein image that has passed the matching authentication and the image quality reaches the preset score is stored in the pre-computing node as the finger vein template of the recognition object, and is preferentially compared with the vein image of the recognition object next time The finger vein template in the front-end computing node performs feature matching to obtain the final authentication result.
The fast authentication method based on finger vein biometrics according to claim 7, wherein when storing the vein image of any identification object collected in real time in the front computing node, further comprising:

Judging whether the number of finger vein templates of the identified object in the front-end computing node reaches a threshold, and if it exceeds the threshold, only keep a specified number of finger vein templates with the highest image quality score and the latest acquisition time.
An electronic device, characterized in that it comprises a processor, a memory, and a computer program stored on the memory and operable on the processor, and when the processor executes the computer program, claims 1 to 1 are realized. 8 Any one of the fast authentication methods based on finger vein biometric technology.
A storage medium, characterized in that a computer program is stored thereon, and when the computer program is executed, the fast authentication method based on finger vein biometric technology described in any one of claims 1 to 8 is realized.