Cross-area finger vein recognition method and system based on connectivity-preserving matching
Through the method of maintaining matching based on connectivity, the different problems of imaging window size and image area caused by the equipment update in finger vein recognition technology are solved, and the accurate matching and recognition of finger vein images is achieved in different areas, improving the accuracy of identity recognition and stability.
Patent Information
- Application Number
- CN202310927272.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-07-26
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2043-07-26
Smart Images

Figure CN116844193B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of finger vein recognition, and in particular, relates to a cross-area finger vein recognition method and system based on connectivity-maintaining matching. Background Art
[0002] Finger vein recognition is a biometric identification technology that uses the veins on the ventral side of the finger for identity recognition. Compared with fingerprint recognition, face recognition, etc., finger vein recognition has some unique advantages, such as liveness recognition: finger vein images can only be collected when the finger is alive; high security: veins are internal features of the finger, and it is difficult to be stolen and copied. From home access control to corporate attendance, to port clearance, biometric identification such as finger vein recognition has become an important technology for personal identity authentication and public safety prevention.
[0003] The inventor found that most of the existing finger vein recognition research work is based on images from a single acquisition device, and the effective area of the registered image and the test image is the same, without considering the replacement of the acquisition device and the difference in the imaging window size of different acquisition devices; in fact, the current electronic products are updated faster and faster, and the replacement of finger vein imaging equipment and recognition equipment is inevitable. Different types of acquisition devices often have large differences in acquisition window size; for example, the image acquisition window length of a finger vein smart door lock is about 52mm, while the acquisition window length of a finger vein recognition module is about 30mm. Assuming that the acquisition device currently used in the recognition system can collect veins at the first and second knuckles of the finger, the acquisition device after the future upgrade can only collect the finger vein at the first knuckle of the finger, which brings about the problem of cross-area finger vein recognition, that is, the identity recognition problem when the effective area of the finger in the registered finger vein image and the test finger vein image is different. Summary of the invention
[0004] In order to solve the above problems, the present invention proposes a cross-area finger vein recognition method and system based on connectivity-preserving matching, which fully considers the different imaging window sizes and the different effective areas of fingers in the image caused by the upgrading of acquisition equipment, and realizes the recognition between images with different areas based on detail point matching; the cross-area finger vein recognition method based on connectivity-preserving matching comprehensively considers the position, descriptor, adjacent points and connectivity of detail points, and can well realize the matching of detail points, thereby realizing identity recognition.
[0005] In order to achieve the above object, the present invention is implemented through the following technical solutions:
[0006] In a first aspect, the present invention provides a cross-area finger vein recognition method based on connectivity-preserving matching, comprising:
[0007] Acquire detail points of a first finger vein image and a second finger vein image, wherein an area of the first finger vein image is smaller than an area of the second finger vein image;
[0008] Determine the descriptors of the detail points, and determine the initial matching point pair set of the first finger vein image and the second finger vein image by judging the distance of the descriptors;
[0009] Using the positions of the matching point pairs in the initial matching point pair set, locating a candidate area similar to the first finger vein image from the second finger vein image;
[0010] The similarity between the first finger vein image and the candidate region is calculated based on the connectivity-maintaining matching, so as to achieve recognition between the first finger vein image and the second finger vein image.
[0011] Furthermore, a single-pixel-wide vein network is extracted from the first finger vein image and the second finger vein image respectively to obtain detail points in the vein network, where the detail points include bifurcation points, endpoints, and directional inflection points.
[0012] Furthermore, a circular neighborhood based on a preset radius is defined for each detail point, and the curvature values of all pixels in the circular neighborhood are used as descriptors of the detail point.
[0013] Furthermore, the Euclidean distances of all detail points in the first finger vein image and all detail points in the second finger vein image are calculated; if the ratio of the minimum value to the second minimum value in the Euclidean distance is less than a predefined threshold, the two detail points corresponding to the minimum value in the Euclidean distance match.
[0014] Furthermore, for a pair of matching detail points, the distance from the detail point in the first finger vein image to the four boundaries of the first finger vein image is calculated; a rectangular area is taken with another detail point as the center in the second finger vein image; the distance from the detail point in the second finger vein image to the four sides of the rectangular area is equal to the distance from the detail point in the first finger vein image to the four boundaries of the first finger vein image, and the rectangular area is the candidate area.
[0015] Furthermore, in the connectivity preserving matching, the initial matching point pairs involved in the non-candidate area are eliminated; and through the neighbor consistency constraint and the connection consistency constraint, the incorrectly matched detail point pairs are further eliminated; the ratio of the number of correctly matched detail points to the number of all detail points is used as the similarity between the first finger vein image and the candidate area, so as to realize the recognition between the first finger vein image and the second finger vein image.
[0016] In a second aspect, the present invention further provides a cross-area finger vein recognition system based on connectivity-preserving matching, comprising:
[0017] A minutiae extraction module is configured to: obtain minutiae points of a first finger vein image and a second finger vein image, wherein an area of the first finger vein image is smaller than an area of the second finger vein image;
[0018] A descriptor determination module is configured to: determine the descriptor of the detail point, and determine the initial matching point pair set of the first finger vein image and the second finger vein image by judging the distance of the descriptor;
[0019] A candidate region module is configured to locate a candidate region similar to the first finger vein image from the second finger vein image using the positions of the matching point pairs in the initial matching point pair set;
[0020] The recognition module is configured to: maintain matching based on connectivity, calculate the similarity between the first finger vein image and the candidate area, and realize recognition between the first finger vein image and the second finger vein image.
[0021] In a third aspect, the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the cross-area finger vein recognition method based on connectivity-preserving matching described in the first aspect.
[0022] In a fourth aspect, the present invention further provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of the cross-area finger vein recognition method based on connectivity-maintaining matching described in the first aspect are implemented.
[0023] Compared with the prior art, the present invention has the following beneficial effects:
[0024] 1. The present invention first defines detail points and corresponding descriptors of two finger vein images of different areas, and then determines the initial matching point pair set of the two finger vein images by judging the distance of the descriptors; uses the position of the matching point pairs in the initial matching point pair set to locate the candidate area similar to the small-area finger vein image from the large-area finger vein image; finally, based on the connectivity-maintaining matching, the similarity between the small-area finger vein image and the candidate area is calculated to achieve recognition between the two finger vein images; compared with the existing finger vein recognition method, the different imaging window sizes and the different effective areas of the fingers in the image caused by the replacement of the acquisition equipment are fully considered, and the recognition between images of different areas is achieved based on detail point matching; the cross-area finger vein recognition method based on connectivity-maintaining matching fully considers the position, descriptor, adjacent points and connectivity of the detail points, and can well achieve the matching of detail points, thereby achieving identity recognition.
[0025] 2. The present invention uses the directional turning points of the finger vein branches as detail points, which effectively increases the number of detail points and enriches the distinguishability of the detail points. BRIEF DESCRIPTION OF THE DRAWINGS
[0026] The drawings in the specification that constitute a part of this embodiment are used to provide a further understanding of this embodiment. The schematic embodiments of this embodiment and their descriptions are used to explain this embodiment and do not constitute improper limitations on this embodiment.
[0027] Figure 1 This is a schematic diagram of the process of Example 1 of the present invention;
[0028] Figure 2 This is the positioning of candidate areas based on matching point pairs according to Embodiment 1 of the present invention. DETAILED DESCRIPTION
[0029] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.
[0030] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present application belongs.
[0031] Embodiment 1:
[0032] This embodiment provides a cross-area finger vein recognition method based on connectivity-preserving matching, including:
[0033] Acquire detail points of a first finger vein image and a second finger vein image, wherein an area of the first finger vein image is smaller than an area of the second finger vein image;
[0034] Determine the descriptors of the detail points, and determine the initial matching point pair set of the first finger vein image and the second finger vein image by judging the distance of the descriptors;
[0035] Using the positions of the matching point pairs in the initial matching point pair set, locating a candidate area similar to the first finger vein image from the second finger vein image;
[0036] The similarity between the first finger vein image and the candidate region is calculated based on the connectivity-maintaining matching, so as to achieve recognition between the first finger vein image and the second finger vein image.
[0037] It can be understood that the first finger vein image and the second finger vein image are respectively a registered finger vein image and a test finger vein image, or a test finger vein image and a registered finger vein image; the first finger vein image can be called a small area image, and the second finger vein image can be called a large area image. Specifically, firstly, the detail points and corresponding descriptors of two finger vein images with different areas are defined, and then the initial matching point pair set of the two finger vein images is determined by judging the distance of the descriptors; the position of the matching point pairs in the initial matching point pair set is used to locate the candidate area similar to the small area finger vein image from the large area finger vein image; finally, based on the connectivity-maintaining matching, the similarity between the small area finger vein image and the candidate area is calculated to realize the recognition between the two finger vein images; compared with the existing finger vein recognition method, the different imaging window sizes caused by the replacement of the acquisition equipment and the different effective areas of the fingers in the image are fully considered, and the recognition between images with different areas is realized based on detail point matching; the cross-area finger vein recognition method based on connectivity-maintaining matching comprehensively considers the position, descriptor, adjacent points and connectivity of detail points, and can well realize the matching of detail points, thereby realizing identity recognition.
[0038] The specific steps of this embodiment are:
[0039] S1. Extraction and characterization of detail points: Extract a single-pixel-wide vein network from the first finger vein image and the second finger vein image, respectively, and obtain detail points in the vein network, where the detail points include bifurcation points, endpoints, and directional inflection points. Define a circular neighborhood based on a preset radius for each detail point, and use the curvature values of all pixels in the circular neighborhood as descriptors of the detail point.
[0040] Specifically, a single-pixel-wide vein network is extracted from the finger vein image using a Gabor filter, and detail points such as bifurcation points, endpoints, and direction inflection points in the vein network are detected. For each detail point, a circular neighborhood with a radius of r and the center of the point is defined, and the curvature of all pixels in the circular neighborhood is used as the descriptor of the detail point.
[0041] S2. Initial matching of detail points: calculating the Euclidean distances of all detail points in the first finger vein image and all detail points in the second finger vein image; for each detail point in the first finger vein image, if the ratio of the minimum value to the second minimum value in the Euclidean distance is less than a predefined threshold, the two detail points corresponding to the minimum value in the Euclidean distance are matched.
[0042] Specifically: Assume that there is a small area image and a large area image to be identified, and the detail point descriptor sets are a i,i=1,2,…,m and b i,i=1,2,…,n, where m and n are the number of detail points in the two images respectively. For the i-th detail point in the small area image, calculate its Euclidean distance with the n detail points in the large area image, that is, Assumption d ij and d ix are the minimum and second minimum values in the distance vector. If the ratio of the minimum value to the second minimum value is less than the predefined threshold t, the i-th detail point in the small area image is considered to match the j-th detail point in the large area image. The above processing is performed on the m detail points in the small area image to obtain the initial matching point pair set of the two images to be matched. Among them, x i ,y i are two-dimensional vectors, representing the spatial position of detail points. i is the detail point in the small area image, y i is the detail point in the large area image, and N is the number of matching point pairs in the two images.
[0043] S3. Positioning of candidate areas: for a pair of matching detail points, calculate the distance from the detail point in the first finger vein image to the four boundaries of the first finger vein image; take a rectangular area with another detail point as the center in the second finger vein image; the distance from the detail point in the second finger vein image to the four sides of the rectangular area is equal to the distance from the detail point in the first finger vein image to the four boundaries of the first finger vein image, and the rectangular area is the candidate area.
[0044] Specifically: Use the position of the detail point pair in the initial matching point pair set to locate the candidate area that is most similar to the small area image from the large area image. i ,y i ), calculate x i The distances to the top, bottom, left, and right boundaries of the small image are [l1, l2, l3, l4]. In the large image, y i Take the rectangular area r as the center i , point y i To the rectangular area r i The distances of the four edges are l1, l2, l3, and l4 respectively. This area is used as a detail point pair (x i ,y i ) is used to calculate the candidate region. Based on the detail point pair (x i ,y i ) is located as follows Figure 2 According to the above steps, N rectangular candidate regions [r1, r2, ..., r N ], and take the union of these candidate regions as the final candidate region r.
[0045] S4, connectivity preserving matching: Connectivity preserving matching is used to eliminate incorrect matching point pairs in the initial matching point pair set to obtain correct matching point pairs, and the ratio of the number of correctly matched detail points to the number of all detail points is used as the similarity between the first finger vein image and the second finger vein image; in the connectivity preserving matching, constraints are defined from three aspects: position preservation, adjacency preservation, and connectivity preservation of detail points to achieve the elimination of incorrect matching point pairs.
[0046] Specifically: In connectivity preserving matching, the initial matching point pairs involved in non-candidate areas are eliminated, and the wrong matching detail point pairs are further eliminated through neighbor consistency constraints and connection consistency constraints, and the ratio of the number of correctly matched detail points to the number of all detail points is used as the similarity between the two images. In connectivity preserving matching, constraints are defined from three aspects: position preservation of detail points (whether the detail points belong to the candidate area), adjacency preservation of detail points, and connectivity preservation of detail points, to achieve the elimination of wrong matching point pairs. In detail,
[0047] S4.1, Position holding:
[0048] In the initial matching point pair, if the detail point in the large area image is not in the candidate area to be located, the relevant matching point pair is considered to be a wrong match and is removed from the initial matching point pair set. The following metric function is defined to measure the position preservation of the detail point pair:
[0049]
[0050] S4.2. Adjacency maintenance:
[0051] In step S4.1, matching point pairs that are not in the candidate area are eliminated, and matching point pairs with different neighborhood structures will be further eliminated. Assume that (x i ,y i )(x j ,y j ) are two pairs of matching points. If x j is x i If the points in the neighborhood are j Also y i Points in the neighborhood. Define the following metric function to measure the adjacency preservation of detail points:
[0052]
[0053]
[0054] Where b∈N a means that b is a point in the neighborhood of a. Indicates that b is not a point in the neighborhood of a.
[0055] S4.3, Connectivity Maintenance:
[0056] The above metric function only considers the position and adjacent relationship of detail points, but does not consider the connection relationship of detail points. i ,y i )(x j ,y j ) are two pairs of matching points. If x j and x i On a vein branch, we hope that y j and i Also on a vein branch. Define the following metric function to measure the connectivity preservation of detail point pairs:
[0057]
[0058]
[0059] Where b∈C a It means that a and b are on a branch of a vein. It means that a and b are not on the same vein branch.
[0060] S4.4. Objective function:
[0061] Taking into account the position preservation, adjacency preservation and connectivity preservation of detail point pairs, the following objective function is defined:
[0062]
[0063] Among them, if the first matching point pair (x i ,y i ) was not eliminated, P i =1; otherwise, P i = 0. Minimizing the above loss can eliminate the wrong matching point pairs and obtain the final set of matching point pairs. The ratio of the number of detail points in the matching point pair set to the number of all detail points in the two images is used as the similarity score between the two images, and the recognition between the images is achieved based on this score.
[0064] Embodiment 2:
[0065] This embodiment provides a cross-area finger vein recognition system based on connectivity-preserving matching, including:
[0066] A minutiae extraction module is configured to: obtain minutiae points of a first finger vein image and a second finger vein image, wherein an area of the first finger vein image is smaller than an area of the second finger vein image;
[0067] A descriptor determination module is configured to: determine the descriptor of the detail point, and determine the initial matching point pair set of the first finger vein image and the second finger vein image by judging the distance of the descriptor;
[0068] A candidate region module is configured to locate a candidate region similar to the first finger vein image from the second finger vein image using the positions of the matching point pairs in the initial matching point pair set;
[0069] The recognition module is configured to: maintain matching based on connectivity, calculate the similarity between the first finger vein image and the candidate area, and realize recognition between the first finger vein image and the second finger vein image.
[0070] The working method of the system is the same as the cross-area finger vein recognition method based on connectivity-preserving matching in Example 1, and will not be repeated here.
[0071] Embodiment 3:
[0072] This embodiment provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps of the cross-area finger vein recognition method based on connectivity-preserving matching described in Embodiment 1 are implemented.
[0073] Embodiment 4:
[0074] This embodiment provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps of the cross-area finger vein recognition method based on connectivity-maintaining matching described in Embodiment 1 are implemented.
[0075] The above description is only a preferred embodiment of the present embodiment and is not intended to limit the present embodiment. For those skilled in the art, the present embodiment may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present embodiment shall be included in the protection scope of the present embodiment.
Claims
1. A cross-area finger vein recognition method based on connectivity-preserving matching, characterized in that: include: Acquire detail points of a first finger vein image and a second finger vein image, wherein an area of the first finger vein image is smaller than an area of the second finger vein image; Determine the descriptor of the detail point, extract the single-pixel-wide vein network from the first finger vein image and the second finger vein image respectively, and obtain the detail points in the vein network, where the detail points include bifurcation points, endpoints, and directional inflection points; define a circular neighborhood based on a preset radius for each detail point, and use the curvature values of all pixels in the circular neighborhood as the descriptor of the detail point; determine the initial matching point pair set of the first finger vein image and the second finger vein image by judging the distance of the descriptors; Using the positions of the matching point pairs in the initial matching point pair set, locating a candidate area similar to the first finger vein image from the second finger vein image; For a pair of matching detail points, the distance from the detail point in the first finger vein image to the four boundaries of the first finger vein image is calculated; a rectangular area is taken with another detail point as the center in the second finger vein image; the distance from the detail point in the second finger vein image to the four sides of the rectangular area is equal to the distance from the detail point in the first finger vein image to the four boundaries of the first finger vein image, and the rectangular area is the candidate area; The similarity between the first finger vein image and the candidate region is calculated based on the connectivity-maintaining matching, so as to achieve recognition between the first finger vein image and the second finger vein image.
2. The cross-area finger vein recognition method based on connectivity-preserving matching as claimed in claim 1, characterized in that: A single-pixel-wide vein network is extracted from the first finger vein image and the second finger vein image respectively, and detail points in the vein network are obtained, where the detail points include bifurcation points, endpoints, and direction inflection points.
3. The cross-area finger vein recognition method based on connectivity-preserving matching as claimed in claim 1, characterized in that: The Euclidean distances of all detail points in the first finger vein image and all detail points in the second finger vein image are calculated; if the ratio of the minimum value to the second minimum value in the Euclidean distance is less than a predefined threshold, the two detail points corresponding to the minimum value in the Euclidean distance match.
4. The cross-area finger vein recognition method based on connectivity-preserving matching as claimed in claim 1, characterized in that: In the connectivity preserving matching, the initial matching point pairs involved in the non-candidate area are eliminated; and the incorrectly matched detail point pairs are further eliminated through the neighbor consistency constraint and the connection consistency constraint; the ratio of the number of correctly matched detail points to the number of all detail points is used as the similarity between the first finger vein image and the candidate area, so as to realize the recognition between the first finger vein image and the second finger vein image.
5. A cross-area finger vein recognition system based on connectivity-preserving matching, characterized in that: include: A minutiae extraction module is configured to: obtain minutiae points of a first finger vein image and a second finger vein image, wherein an area of the first finger vein image is smaller than an area of the second finger vein image; The descriptor determination module is configured to: determine the descriptor of the detail point, extract the single-pixel-wide vein network from the first finger vein image and the second finger vein image respectively, and obtain the detail points in the vein network, where the detail points include bifurcation points, endpoints, and directional inflection points; define a circular neighborhood based on a preset radius for each detail point, and use the curvature values of all pixels in the circular neighborhood as the descriptor of the detail point; and determine the initial matching point pair set of the first finger vein image and the second finger vein image by judging the distance of the descriptors; A candidate region module is configured to locate a candidate region similar to the first finger vein image from the second finger vein image using the positions of the matching point pairs in the initial matching point pair set; For a pair of matching detail points, the distance from the detail point in the first finger vein image to the four boundaries of the first finger vein image is calculated; a rectangular area is taken with another detail point as the center in the second finger vein image; the distance from the detail point in the second finger vein image to the four sides of the rectangular area is equal to the distance from the detail point in the first finger vein image to the four boundaries of the first finger vein image, and the rectangular area is the candidate area; The recognition module is configured to: maintain matching based on connectivity, calculate the similarity between the first finger vein image and the candidate area, and realize recognition between the first finger vein image and the second finger vein image.
6. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the cross-area finger vein recognition method based on connectivity-preserving matching as described in any one of claims 1 to 4 are implemented.
7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the cross-area finger vein recognition method based on connectivity-preserving matching as described in any one of claims 1 to 4 are implemented.
