WO2022110492A1

WO2022110492A1 - Finger vein-based identity identification method and apparatus, computer device, and storage medium

Info

Publication number: WO2022110492A1
Application number: PCT/CN2020/141305
Authority: WO
Inventors: 王丹丹; 王晓亮; 陈�光
Original assignee: 广州广电运通金融电子股份有限公司
Priority date: 2020-11-24
Filing date: 2020-12-30
Publication date: 2022-06-02
Also published as: CN112508064A

Abstract

The present application relates to a finger vein-based identity identification method and apparatus, a computer device, and a storage medium. Said method comprises: acquiring branch information and bifurcation point information in vein skeleton diagrams corresponding to two finger vein images to be matched; respectively performing matching on the branch information and the bifurcation point information of the two vein skeleton diagrams to obtain multiple groups of matched bifurcation point pairs; acquiring coordinate offset amounts of each group of matched bifurcation point pairs, and obtaining multiple groups of target bifurcation point pairs by screening the groups of matched bifurcation point pairs according to the coordinate offset amounts; acquiring a similarity of each group of target bifurcation point pairs, and acquiring branch lengths of all branches in the vein skeleton diagrams; and according to the similarity of each group of target bifurcation point pairs and the branch lengths of all branches in the vein skeleton diagrams, obtaining a similarity of the two finger vein images, and performing identity identification according to the similarity. The accuracy of a result of calculation of a similarity of two finger vein images can be improved by using said method.

Description

Finger vein identification method, device, computer equipment and storage medium

technical field

The present application relates to the technical field of finger veins, and in particular, to a finger vein identification method, device, computer equipment and storage medium.

Background technique

Finger vein biometric technology is a method of identity authentication using the blood vessel distribution pattern formed when blood flows through the superficial blood vessels of the finger as a biometric. Store and then match for personal identification.

When performing feature matching in the existing finger vein identification methods, most of them are based on the fork points or endpoints in the image to complete the identification. However, this method of matching by fork or endpoint has low identification accuracy.

SUMMARY OF THE INVENTION

Based on this, it is necessary to provide a finger vein identification method, device, computer equipment and storage medium in view of the technical problem of low identification accuracy in the above-mentioned finger vein identification method.

A method for identifying a finger vein, the method comprising:

Obtain the branch information and fork point information in the vein skeleton map corresponding to the two finger vein images to be matched;

Match the branch information and fork point information of the two vein skeleton maps respectively to obtain multiple matched pairs of fork points;

Obtain the coordinate offsets of each group of matched fork point pairs, and filter out multiple groups of target fork point pairs from each group of matched fork point pairs according to the coordinate offsets;

Obtain the similarity of each target fork point pair, and obtain the branch lengths of all branches in each of the vein skeleton diagrams; the similarity of the target fork point pair is the matched branch of each group in the target fork point pair The sum of the similarity;

According to the similarity of each target fork point pair and the branch lengths of all branches in each of the vein skeleton images, the similarity of the two finger vein images is obtained, and the identity is identified according to the similarity.

In one of the embodiments, the branch information and fork point information of the two vein skeleton maps are respectively matched to obtain multiple sets of matched fork point pairs, including:

Obtain the coordinate offsets of each group of fork point pairs in the two vein skeleton maps, and select multiple groups of fork point pairs whose coordinate offsets are less than the offset threshold from each group of fork point pairs to form the first fork point. pair collection;

Acquire the main direction angles of the branches corresponding to each group of fork point pairs in the first fork point pair set, and determine the combination relationship between the branches corresponding to each group of fork point pairs to obtain a plurality of target branch combinations; The main direction angle of a branch is the mean value of the direction angles of each pixel in the branch;

Calculate the difference of the main direction angles of the branches under each target branch combination, and from the first set of fork point pairs, filter out multiple groups of forks whose differences in the main direction angles of the branches under the target branch combination are less than the difference threshold point pairs to form a second set of fork point pairs;

Obtain the branch linear correlation coefficients under the target branch combination in each group of fork point pairs in the second fork point pair set, and filter out multiple groups whose linear correlation coefficients are greater than the correlation coefficient threshold from the second fork point pair set A pair of forks, as the matched pair of forks.

In one of the embodiments, the determining the combination relationship between the branches corresponding to each group of fork point pairs obtains a plurality of target branch combinations, including:

Arrange and combine the corresponding branches of each group of fork points to obtain a plurality of branch combinations to be matched;

If any branch corresponds to at least two to-be-matched branch combinations, the difference between the branch main direction angles under each to-be-matched branch combination corresponding to the branch is calculated, and the to-be-matched branch combination with the smallest difference is used as the target branch combination.

In one embodiment, the obtaining the linear correlation coefficients of the lower branches of the target branch combination in each group of fork point pairs in the second fork point pair set includes:

Convert the two-dimensional coordinate vector of the branch corresponding to each group of fork pairs into a one-dimensional coordinate vector;

The correlation coefficient of the one-dimensional coordinate vector of the lower branch of the target branch combination in each group of fork point pairs is calculated as the linear correlation coefficient of the lower branch of the target branch combination.

In one embodiment, obtaining the coordinate offsets of each group of matched fork point pairs, and filtering out multiple groups of target fork point pairs from each group of matched fork point pairs according to the coordinate offsets, including:

According to the coordinate offset, each group of matched fork-point pairs is divided into a plurality of fork-point pair sets;

Obtain the similarity of each group of matched fork-point pairs, calculate the sum of the similarities of the fork-point pairs in each fork-point pair set, and obtain the cumulative sum of the similarity of each fork-point pair set;

Each group of matched fork-point pairs included in the fork-point pair set with the maximum similarity sum is taken as the target fork-point pair.

In one embodiment, the similarity of the two finger vein images is obtained according to the similarity of each target fork point pair and the branch lengths of all branches in each of the vein skeleton images, including:

The similarity of each of the target fork pairs is added to obtain the sum of the similarity, and the branch lengths of all branches are added to obtain the sum of the branch lengths;

Calculate the ratio between the sum of the similarity and the sum of the branch lengths, and multiply the ratio by 2 to obtain a value as the similarity of the two finger vein images.

In one embodiment, after filtering out multiple groups of target fork point pairs from each group of matched fork point pairs according to the coordinate offset, the method further includes:

Determine the unmatched fork pairs in the matched branches in each group of target fork pairs as the fork pairs to be matched;

Identifying whether the pair of fork points to be matched matches, and identifying whether the branches under each target branch combination corresponding to the pair of fork points to be matched match, if both match, adding the pair of fork points to be matched in the set formed by the target fork pair.

In one embodiment, the determining of unmatched branches in each group of target fork point pairs, and obtaining the unmatched fork point pairs in the unmatched branches, as the fork point pairs to be matched, includes:

When there is an unmatched branch end point in each group of matched branches in the set formed by the target fork point pair, calculate the branch length difference between the two branches corresponding to the unmatched branch end point;

If the branch length difference is not greater than the length difference threshold, the two unmatched branch end points are used as the pair of fork points to be matched;

If the branch length difference is greater than the length difference threshold, a pseudo fork point is added to the longer branch corresponding to the end point of the unmatched branch, and the end point of the shorter branch and the pseudo fork point are used as the waiting point. Match fork pairs.

A finger vein identification device, the device comprises:

a feature acquisition module, used for acquiring branch information and fork point information in the vein skeleton map corresponding to the two finger vein images to be matched;

The feature matching module is used to match the branch information and fork point information of the two vein skeleton maps respectively to obtain multiple sets of matched fork point pairs;

A cross-point pair screening module, used to obtain the coordinate offsets of the matched cross-point pairs of each group, and screen out multiple groups of target cross-point pairs from the matched cross-point pairs according to the coordinate offsets;

a length obtaining module, used to obtain the similarity of each target fork point pair, and obtain the branch lengths of all branches in each of the vein skeleton diagrams; the similarity of the target fork point pair is the target fork point pair The sum of the similarity of the matched branches in each group;

The similarity determination module is configured to obtain the similarity of the two finger vein images according to the similarity of each target fork point pair and the branch lengths of all the branches, and perform identity recognition according to the similarity.

A computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the following steps when executing the computer program:

A computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

The above-mentioned finger vein identification method, device, computer equipment and storage medium, by acquiring the branch information and fork point information in the vein skeleton map corresponding to the two finger vein images to be matched, so that the branches of the two finger vein images can be respectively identified. Information and fork point information are matched to obtain multiple sets of matched fork point pairs. According to the coordinate offset of each group of matched fork point pairs, the target fork point pairs are screened from each group of matched fork point pairs as the final match the benchmark. Finally, according to the obtained similarity of each group of target fork point pairs and the branch lengths of all branches in the two vein skeleton images, the similarity of the two finger vein images is calculated, and the identity is identified according to the similarity. This method not only obtains the fork point information, but also obtains the branch information, and calculates the similarity of the two finger vein images through the similarity of the corresponding branches of each fork point, so as to avoid the similarity only by the number of fork point pairs. The problems of easy errors and low accuracy in the degree calculation improve the accuracy of the calculation results of the similarity of the two finger vein images. The influence on the result of similarity calculation by the cross point of matching error is avoided, the accuracy of the calculation result of similarity is further improved, and thus the reliability of the identification result is improved.

Description of drawings

1 is a schematic flowchart of a finger vein identification method in one embodiment;

Figure 2 (a) and Figure 2 (b) are respectively the direction angle corresponding to the position of the pixel point adjacent to the pixel X and the pixel X in one embodiment;

Fig. 2 (c) is the schematic diagram of the direction angle of each pixel point in the branch in one embodiment;

3 is a schematic diagram of a branch and a fork in one embodiment;

4 is a schematic diagram of a vein skeleton diagram in one embodiment;

5 is a schematic diagram of a two-dimensional histogram of coordinate offsets in one embodiment;

6 is a schematic flowchart of the steps of obtaining a matched pair of forks in one embodiment;

7 is a schematic diagram of a pair of fork points to be matched in one embodiment;

8 is a schematic flowchart of expanding fork pairs to be matched in one embodiment;

9 is a schematic flowchart of a step of determining a point pair to be matched in one embodiment;

10 is a schematic diagram of expanding point pairs to be matched in one embodiment;

11 is a structural block diagram of a finger vein identification device in one embodiment;

Figure 12 is a diagram of the internal structure of a computer device in one embodiment.

Detailed ways

In order to make the purpose, 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 embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, but not to limit the present application.

In one embodiment, as shown in FIG. 1 , a finger vein identification method is provided. In this embodiment, the method is applied to a terminal for illustration. It can be understood that the method can also be applied to a server, or It is applied to a system including a terminal and a server, and is realized through the interaction between the terminal and the server. In this embodiment, the method includes the following steps:

Step S102, acquiring branch information and fork point information in the vein skeleton map corresponding to the two finger vein images to be matched.

The branch information includes attribute information such as branch identification, branch length, branch main direction angle, start point, end point, and coordinates.

where a branch represents an ordered combination of pixels connecting two forks or endpoints.

Wherein, the branch identifier is a flag representing the uniqueness of the branch.

Among them, the branch length represents the number of pixels forming the branch.

Wherein, the branch coordinates represent the two-dimensional coordinate (x, y) array of each pixel point constituting the branch.

Among them, the starting point and the ending point of the branch are the two endpoints of the branch. If one of them is selected as the starting point, the other one is the ending point.

Among them, the main direction angle of the branch represents the average value of the direction angle of each pixel point of the branch, as shown in Fig. 2(a) and Fig. 2(b), respectively, the position of the pixel adjacent to the pixel X and the direction corresponding to the pixel X As shown in Figure 2(c), it is a schematic diagram of the direction angle of each pixel point in the branch, and the direction angle of each pixel point is determined according to the position of the current pixel point and the next adjacent pixel point, for example, if the next pixel point If the point is at position 3 in Fig. 2(a), the direction angle of pixel X is -45°, and if the next pixel is at position 7 in Fig. 2(a), the direction angle of pixel X is 135°. Thus, as shown in Figure 2(c), the orientation angle of each pixel in a branch can be obtained, and the main orientation angle of the branch can be obtained by calculating the mean value of the orientation angle of each pixel.

The cross point information includes information such as cross point identification, cross point coordinates, number of connected branches, attributes of each connected branch, and the like.

Among them, the cross point represents the pixel point where two or more branches are cross-connected.

Wherein, the fork point identifier is a sign indicating the uniqueness of the fork point.

The cross point coordinates are the two-dimensional coordinates of the cross point pixels, which can be represented by (x, y).

As shown in Figure 3, it is a schematic diagram of a branch and a fork point. In the figure, the pixel point P is a fork point, and b1, b2 and b3 are three branches connected to the fork point P, and the fork point P can be regarded as b1, b2 and b3 The pixels where the three branches are cross-connected.

In the specific implementation, the finger vein image can be collected by the finger vein acquisition device, and after the finger vein image is obtained, the finger vein image is preprocessed to obtain the vein skeleton map, and the skeleton map features are extracted from the vein skeleton map, wherein the skeleton map Features include branch information and fork information. More specifically, the image preprocessing steps of the finger vein image include: locating the finger rectangle region (ROI), correcting the rotation angle of the finger plane, performing preprocessing such as grayscale processing and size normalization on the image, and obtaining the size, grayscale, etc. The normalized image is further processed by vein enhancement, binarization and skeleton refinement to obtain the vein skeleton map, and the feature extraction operation is performed on the vein skeleton map to obtain the branch information and fork in the vein skeleton map. point information. As shown in FIG. 4 , it is a schematic diagram of a vein skeleton diagram, including features such as several fork points and multiple branches connected to the fork points.

In step S104, the branch information and the fork point information of the two vein skeleton maps are respectively matched to obtain a plurality of matched pairs of fork points.

In the specific implementation, the feature matching of the two vein skeleton maps is the matching of the fork points and the matching of the branches. The matching of the fork points can be obtained by obtaining the coordinates of each fork point, and determining whether the fork points match according to the coordinates. The main direction angle of each branch corresponding to the fork point is used to determine whether the two fork points match according to the main direction angle of the branch. The matching of the branches can be determined by calculating the linear correlation coefficient of the two branches, and determining whether the two branches match according to the linear correlation coefficient. When matching the branches corresponding to the two fork points, the combination relationship between the branches can be determined first according to the main direction angle of each branch, and the linear correlation coefficient between the two branches under the target combination can be determined. The linear correlation coefficient of , judges whether the two branches under the target combination match. Therefore, through the matching between each fork point and branch, multiple sets of matched fork point pairs can be obtained, wherein each fork point pair is composed of a fork point in the first vein skeleton diagram and a fork point in the second vein skeleton diagram. consisting of a fork point.

In step S106, the coordinate offsets of each group of matched fork point pairs are acquired, and a plurality of groups of target fork point pairs are screened from each group of matched fork point pairs according to the coordinate offsets.

The coordinate offset represents the difference between the coordinates of the two cross points. Since the cross point coordinates are two-dimensional vector coordinates, the coordinate offset may include the difference between the coordinates in the x-direction and the coordinate in the y-direction.

In the specific implementation, since there may be incorrectly matched fork pairs among the multiple sets of matched fork pairs obtained through fork and branch matching, screening processing is required to screen out the most consistent fork pairs from each group of matched fork pairs. Several sets of good target fork point pairs are used as the final matching benchmark for the two finger vein images.

Further, step S106 specifically includes: dividing each group of matched fork point pairs into a plurality of fork point pair sets according to the coordinate offset; The sum of the similarities of the fork-point pairs is obtained, and the cumulative sum of the similarity of each fork-point pair set is obtained; the matched fork-point pairs included in each group of fork-point pairs included in the set of fork-point pairs with the maximum similarity sum are taken as the target fork-point pair .

Specifically, the coordinates of each fork point in each group of matched fork point pairs can be obtained, the coordinate offset of each group of matched fork point pairs can be calculated, and the coordinate offset of each group of matched fork point pairs can be counted using a two-dimensional histogram According to the coordinate offset, each group of matched fork pairs is grouped, and the offsets are divided into a set of fork pairs, thereby obtaining multiple sets of fork pairs, wherein the obtained set of fork pairs is The total number does not exceed 3. For example, as shown in Figure 5, which is a schematic diagram of a two-dimensional histogram of coordinate offsets, 6, 3, 12, 16, 17, and 20 in the middle of the figure are all distributed around 12, indicating that the offsets are consistent, then Its corresponding fork-point pair can be regarded as a fork-point pair set. After obtaining multiple sets of fork-point pairs, calculate the similarity of each group of fork-point pairs in each set of fork-point pairs, add the similarities of each group of fork-point pairs, and obtain the cumulative sum of the similarities of the set of fork-point pairs. The set of fork point pairs with the maximum accumulated similarity is used as the final matching reference point set, and each group of matched fork point pairs in the final matching reference point set is used as the target fork point pair.

Step S108, obtaining the similarity of each target fork point pair, and obtaining the branch lengths of all branches in each vein skeleton diagram; the similarity degree of the target fork point pair is the sum of the similarities of each group of matched branches in the target fork point pair .

The similarity of the matched branches of each group is obtained by multiplying the matching length of the matched branches of the group and the linear correlation coefficient of the matched branches of the group.

In the specific implementation, the two-dimensional vector of the coordinates of each branch in the target fork point pair can be converted into a one-dimensional vector, and the linear correlation coefficient of each group of matching branches in each group of target fork point pairs can be calculated. The product of the length and the linear correlation coefficient of the corresponding groups of matching branches is used as the similarity of each group of matching branches, and then the sum of the similarity of each group of target forks to the corresponding groups of matching branches is calculated as a group of target forks Point-to-point similarity. For example, the similarity of the two branches bi and cj can be expressed as: Sb=ρ _i,j ×l _i,j , where ρ _i,j represents the linear correlation coefficient of the branch bi and the branch cj, and li _,j represents the branch The matching length of bi and branch cj. Then the similarity of the target fork point pair can be expressed as:

In step S110, the similarity of the two finger vein images is obtained according to the similarity of each target fork point pair and the branch lengths of all branches in each vein skeleton image, and the identity is identified according to the similarity.

In the specific implementation, after obtaining the similarity of each group of target fork-point pairs, the similarity of each group of target fork-point pairs is added to obtain the sum of the similarity, and the branch lengths of all branches in the two vein skeleton diagrams are calculated. Add up to obtain the sum of branch lengths, and further calculate the ratio of the sum of the similarity of each target fork point pair to the sum of the branch lengths of all branches.

In practical applications, since the ratio is actually the ratio of the length of all branches in the two vein skeleton maps after multiplying the matching length of the matching branch and the linear correlation coefficient, the matching length of each group of matching branches is calculated only once, so that The ratio of the sum of the similarity of each target fork point pair to the sum of the branch lengths of all branches is less than 0.5. Therefore, in order to facilitate identification, the ratio can be multiplied by 2, and 1 is used as the benchmark for matching two finger vein images. Then the similarity of two finger vein images can be expressed as:

Among them, Sb _k represents the similarity of the k-th pair of matching branches, l _i represents the branch length of the i-th branch in the first vein skeleton map, and l _j represents the branch length of the j-th branch in the second vein skeleton map , N represents the total number of matching branches in the two vein skeleton maps, m represents the total number of branches in the first vein skeleton map, n represents the total number of branches in the second vein skeleton map, N<min {m,n}.

In the above finger vein identification method, the branch information and fork point information in the vein skeleton map corresponding to the two finger vein images to be matched are obtained, so as to match the branch information and fork point information of the two finger vein images respectively. , obtain multiple sets of matched fork-point pairs, and filter out target fork-point pairs from each group of matched fork-point pairs according to the coordinate offsets of each group of matched fork-point pairs, as the final matching reference. Finally, according to the obtained similarity of each group of target fork point pairs and the branch lengths of all branches in the two vein skeleton images, the similarity of the two finger vein images is calculated, and the identity is identified according to the similarity. This method not only obtains the fork point information, but also obtains the branch information, and calculates the similarity of the two finger vein images through the similarity of the corresponding branches of each fork point, so as to avoid the similarity only based on the number of fork point pairs. It is easy to make mistakes and the accuracy is not high in the degree calculation, thereby improving the accuracy of the calculation results of the similarity of the two finger vein images, and by screening the obtained matching fork pairs according to the coordinate offset , which further avoids the influence of mismatched cross points on the similarity calculation result, improves the accuracy of the similarity calculation result, and thus improves the credibility of the identity recognition result.

In one embodiment, as shown in FIG. 6 , the foregoing step S104 specifically includes:

Step S602: Obtain the coordinate offsets of each group of fork point pairs in the two vein skeleton maps, and select multiple groups of fork point pairs whose coordinate offsets are less than the offset threshold from each group of fork point pairs to form a first fork. Point-to-point collection.

Wherein, each set of fork point pairs is composed of a fork point in the first vein skeleton map and a fork point in the second vein skeleton map.

In a specific implementation, the step of acquiring the coordinate offsets of each group of fork point pairs in the two vein skeleton maps includes: acquiring the coordinates of each fork point in the first vein skeleton map to form a first coordinate set, and acquiring a second set of coordinates The coordinates of each fork point in the vein skeleton diagram form a second coordinate set; traverse and calculate the difference between each first coordinate in the first coordinate set and each second coordinate in the second coordinate set, and obtain multiple sets of fork point pairs The coordinate offset of . The following takes the matching of the fork point P and the fork point Q in FIG. 7 as an example to describe the fork point matching method. The coordinate matching of the fork point P and the fork point Q can be expressed by the following relational expression, the fork point P and the fork point The coordinate offsets of Q in the x-direction and the y-direction are compared with the corresponding offset thresholds respectively. When the coordinate offset in the direction is less than the second offset threshold, the fork point pair composed of the fork point P and the fork point Q is screened out, and so on, the cross point pairs in the x direction and the y direction are screened out from each group of fork point pairs. A plurality of sets of fork point pairs whose coordinate offsets are all less than the corresponding offset thresholds form a first set of fork point pairs. Among them, the absolute value of the coordinate offset is compared with the offset threshold. The first offset threshold in the x-direction and the second offset threshold in the y-direction may or may not be equal.

|P.x-Q.x|＜T.x

|P.y-Q.y|＜T.y

Among them, Px and Qx represent the coordinates of the fork points P and Q in the x direction respectively, Py and Qy represent the coordinates of the fork points P and Q in the y direction, respectively, and Ty and Ty represent the offset thresholds in the x and y directions, respectively.

Step S604: Obtain the main direction angles of the branches corresponding to each group of fork point pairs in the first fork point pair set, and determine the combination relationship between the branches corresponding to each group of fork point pairs to obtain multiple target branch combinations; The main direction angle of a branch is the average value of the direction angles of each pixel point in the branch; the difference between the main direction angles of the branches under each target branch combination is calculated, and the target branch combination is selected from the first set of fork point pairs. A plurality of sets of fork point pairs whose differences in the main direction angles of the branches are smaller than the difference threshold form a second set of fork point pairs.

In the specific implementation, as shown in the fork points P and Q in FIG. 7 , there may be multiple branches corresponding to each fork point. Therefore, it is necessary to first determine the combination relationship between the branches in each group of fork point pairs to obtain multiple The target branch combination is to calculate the difference between the main direction angles of the two branches under each target branch combination, and from the first set of fork point pairs, filter out the main direction angle differences of the corresponding target branch combinations. A plurality of sets of fork point pairs with a threshold value are formed to form a second set of fork point pairs. For example, if the target branch combination of the fork-point pair composed of the fork-point P and the fork-point Q in Fig. 7 is <b1, c1>, <b2, c2> and <b3, c3>, as shown in the following relational expression, then When the difference between the main direction angle P.θ1 of the branch b1 of the fork point P and the main direction angle Q.θ1 of the branch c1 of the fork point Q is less than the difference threshold T.θ, and the main direction of the branch b2 of the fork point P The difference between the angle P.θ2 and the main direction angle Q.θ2 of the branch c2 of the fork point Q is smaller than the difference threshold T.θ, and the main direction angle P.θ3 of the branch b3 of the fork point P and the branch c3 of the fork point Q When the difference of the main direction angle Q. θ3 of , is smaller than the difference threshold T. θ, the fork point P and the fork point Q may be added to the second fork point pair set. The absolute value of the difference between the main direction angles of the two branches is compared with the difference threshold.

|P.θ1-Q.θ1|＜T.θ

|P.θ2-Q.θ2|＜T.θ

|P.θ3-Q.θ3|＜T.θ

Step S606, obtaining the branch linear correlation coefficients under the target branch combination in each group of fork point pairs in the second fork point pair set, and screening out multiple groups of fork point pairs whose linear correlation coefficients are greater than the correlation coefficient threshold from the second fork point pair set , as matched pairs of forks.

In the specific implementation, the target branch combinations in the fork point P and the fork point Q in Figure 7 are <b1, c1>, <b2, c2> and <b3, c3>, respectively calculate <b1, c1>, <b2, The linear correlation coefficients of the branches under the three combinations c2> and <b3, c3>, when the linear correlation coefficients of the branches under each combination are greater than the threshold of the correlation coefficient, the fork point P and the fork point Q are used as the matching fork point pair, the The relational expression can be expressed as ρ _i,j >T.ρ; i∈[1,n],j∈[1,m], where ρ _i,j represent the linear correlation coefficient between branch bi and branch bj, T.ρ Represents the correlation coefficient threshold.

Further, in one embodiment, obtaining the linear correlation coefficients of the branches under the target branch combination in each group of fork point pairs in the second fork point pair set includes: transforming the two-dimensional coordinate vectors of the branches corresponding to each group of fork point pairs is a one-dimensional coordinate vector; calculate the correlation coefficient of the one-dimensional coordinate vector of the lower branch of the target branch combination in each group of fork point pairs, as the linear correlation coefficient of the lower branch of the target branch combination.

Specifically, since each branch has a combination of multiple pixel points, the obtained branch coordinates are an array composed of two-dimensional coordinates of pixel points, and each branch coordinate is converted from a two-dimensional vector into a one-dimensional vector, that is, a two-dimensional vector The array is converted to a one-dimensional vector array. For example, as shown in Figure 7, if the two-dimensional coordinate vector of branch b1 is recorded as [(x ₀ , y ₀ ), (x ₁ , y ₁ ), (x ₂ , y ₂ )...(x _n-1 , y _n-1 )], convert it into a one-dimensional coordinate vector to obtain X=(x ₀ , y ₀ , x ₁ , y ₁ , . . . x _n-1 , y _n-1 ). Similarly, the two-dimensional coordinate vector of the branch c1 corresponding to the branch b1 can be converted into a one-dimensional coordinate vector Y. Then the linear correlation coefficient of the target branch combination <b1, c1> can be converted into the linear correlation coefficient of the vectors X and Y, and the linear correlation coefficient of the branch under the target branch combination can be calculated by the following relational formula. By calculating the linear correlation coefficient of the branch under the target branch combination in each group of fork point pairs, it is convenient to screen each group of fork point pairs according to the linear correlation coefficient.

COV(X,Y)=E(XY)-E(X)E(Y)

In this embodiment, the first screening is performed with the coordinate offset with the smallest amount of calculation, the second screening is performed with the difference between the main direction angles between the combined branches with the smallest amount of calculation, and finally the amount of calculation is passed. The linear correlation coefficient of the branch with the largest target combination is screened for the third time, and multiple sets of matching fork-point pairs are obtained. According to the calculation amount from small to large, the method of layer-by-layer screening can reduce the calculation of the matching process of each group of fork point pairs. It can improve the screening efficiency, thereby improving the acquisition rate of the similarity between the two finger vein images to be matched.

In one embodiment, in the above step S604, the combination relationship of each group of fork points to the corresponding branches is determined, and the step of obtaining a combination of multiple target branches specifically includes: arranging and combining the corresponding branches of each group of fork points, Obtain a plurality of branch combinations to be matched; if any branch corresponds to at least two branch combinations to be matched, calculate the difference between the main direction angles of the branches under each branch combination to be matched corresponding to the branch, and select the branch with the smallest difference to be matched. Branch combination, as the target branch combination.

Wherein, each branch combination to be matched is composed of a branch corresponding to the first fork point and a branch corresponding to the second fork point in a group of fork point pairs.

In the specific implementation, taking FIG. 7 as an example, all the branches of the fork point P and all the branches of the fork point Q can be arranged and combined, and each branch of the fork point P corresponds to a plurality of branch combinations to be matched. The branch b1 of point P corresponds to three branch combinations to be matched, <b1, c1>, <b1, c2> and <b1, c3>, and the difference between the main direction angles of the two branches under various branch combinations to be matched is calculated, The branch combination with the smallest difference is used as the target branch combination. For example, if the difference between the main direction angles of the branch b1 and the branch c1 is the smallest, <b1, c1> is used as the target branch combination, indicating that the branch b1 and the branch c1 are the optimal branch combination. Therefore, it can be obtained that the optimal branch combination modes of the branches corresponding to the fork points P and Q are <b1, c1>, <b2, c2> and <b3, c3>.

In this embodiment, by arranging and combining the corresponding branches of each group of fork points, a plurality of branch combinations to be matched are obtained, and multiple groups of target branch combinations are selected from each branch combination to be matched according to the difference of the main direction angles of the branches, The combination relationship between the branches corresponding to each group of fork point pairs is obtained, so that each group of fork point pairs can be screened according to the combination relationship to obtain matching fork point pairs.

In one embodiment, the above step S110 specifically includes: adding the similarities of each target fork point pair to obtain the sum of the similarities, and adding the branch lengths of all branches to obtain the sum of the branch lengths; calculating the similarity The ratio of the sum to the sum of the branch lengths, and the result obtained by multiplying the ratio by 2 is used as the similarity of the two finger vein images.

In this embodiment, the matching length of the matching branches in each group of target fork point pairs in the two finger vein images is multiplied by the linear correlation coefficient to obtain the similarity of the target fork point pair, and the similarity of each target fork point pair is calculated 2 times the ratio of the sum to the sum of the branch lengths of all branches, as the similarity of the two finger vein images, the finger vein images are matched by introducing the branch length, branch similarity, etc. Matching can effectively improve the accuracy of matching results.

In one embodiment, after the above step S106, the method further includes: determining the unmatched fork pairs in the matched branches in each group of target fork pairs as the fork pairs to be matched; identifying whether the fork pairs to be matched are not matched matching, and identifying whether the branches under each target branch combination corresponding to the pair of forks to be matched match, and if both match, the pair of forks to be matched is added to the set formed by the pair of target forks.

In the specific implementation, after the preliminary fork point pair matching is performed, there may be some matching fork point pairs that have not been identified. After matching, it can also identify whether there is an unmatched end point in each group of matching branches in each group of target fork point pairs, and search for unmatched branches and unmatched branches whether there is an unmatched fork point pair, as the fork to be matched. The fork matching method and the branch matching method described in the above embodiments are used to judge whether the fork pair to be matched matches, and whether the branches under each target branch combination corresponding to the fork pair to be matched match. If both of them match, the pair of fork points to be matched is added to the set formed by the pair of target fork points.

In this embodiment, the unmatched fork pair and the unmatched branch in the target fork pair are expanded, and when it is judged that the unexpanded fork pair is a matching point pair, it is added to the target fork pair In the formed set, the calculation of the similarity of the two finger vein images is involved, which further improves the reliability and accuracy of the calculation result of the similarity of the finger vein images.

In one embodiment, after determining the unmatched fork pairs in the matched branches in each group of target fork pairs as the fork pairs to be matched, the method further includes: adding the fork pairs to be matched to the fork points to be matched For the set, identify whether each fork pair to be matched matches in the set of fork pairs to be matched; as shown in FIG. Add unmatched fork pairs to the set of fork pairs to be matched. When matching each group of unmatched fork pairs in the set of fork pairs to be matched, first determine whether the set of fork pairs to be matched is empty, that is, to determine whether the fork pairs to be matched are greater than 0, and when the fork pairs to be matched are matched When it is greater than 0, the unmatched fork pairs in each group in the set are judged one by one to be matched, until the traversal ends.

In one embodiment, as shown in FIG. 9 , the determining of the unmatched fork pairs in the matched branches in each group of target fork pairs, as the fork pairs to be matched, specifically includes:

Step S902, when there is an unmatched branch end point in each group of matched branches in the set formed by the target fork point pair, calculate the branch length difference of the two matching branches corresponding to the unmatched branch end point;

Step S904, if the branch length difference is not greater than the length difference threshold, the end points of the two unmatched branches are used as the pair of forks to be matched;

Step S906, if the branch length difference is greater than the length difference threshold, a pseudo-fork point is added to the longer branch in the matching branch corresponding to the end point of the unmatched branch, and the end point and the pseudo-fork point of the shorter branch are used as the fork to be matched. Right.

In the specific implementation, taking the schematic diagram of the extended point pair to be matched as shown in FIG. 10 as an example, this embodiment will be described. The cross points P1 and Q1 in the figure are the matched reference point pairs, that is, the target cross point pair, < b1, c1>, <b2, c2> and <b3, c3> are three sets of matching branches. It can be seen that there are unmatched branch end points in the matched branches <b1,c1> and <b3,c3> in the target fork point pair composed of P1 and Q1. Then calculate the branch length difference between the two matching branches b1 and c1 corresponding to the unmatched branch end points, and the branch length difference between the matching branches b3 and c3. Since the branch length difference between b1 and c1 shown in FIG. 10 is less than the length difference threshold, the two unmatched end points P2 and Q2 in b1 and c1 can be used as the pair of fork points to be matched and added to the set of fork point pairs to be matched. . And the branch length difference between b3 and c3 is greater than the length difference threshold, then a pseudo-fork point Q3 is added to the longer branch c3, and the end point P3 and the pseudo-fork point Q3 of the shorter branch b3 are used as the pair of fork points to be matched. In the set of fork pairs to be matched.

In this embodiment, when there is an unmatched branch end point in each group of matched branches corresponding to the target fork point pair, by calculating the branch length difference of the two matching branches, according to the comparison between the branch length difference and the length difference threshold, in the branch When the length difference is greater than the length difference threshold, the extension of the pair of points to be matched is realized by adding a pseudo cross point, which avoids directly using the end points of the two matching branches as the pair of points to be matched. Since the coordinates of the two end points are far apart, the result obtained is: The two end points do not match, which wastes computing resources and time, and also misses a pair of possible matching fork points, thus affecting the accuracy of the matching results of the two finger vein images.

It should be understood that although the steps in the flowcharts of FIGS. 1 , 6 and 9 are sequentially displayed in accordance with the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, the execution of these steps is not strictly limited to the order, and these steps may be performed in other orders. Moreover, at least a part of the steps in FIG. 1 , FIG. 6 and FIG. 9 may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but may be executed at different times. The order of execution of the sub-steps or phases is also not necessarily sequential, but may be performed alternately or alternately with other steps or at least a portion of the sub-steps or phases of the other steps.

In one embodiment, as shown in FIG. 11 , a finger vein identification device is provided, including: a feature acquisition module 1102 , a feature matching module 1104 , a cross-point pair screening module 1106 , a length acquisition module 1108 and a similarity determination module 1110, where:

A feature acquisition module 1102, configured to acquire branch information and fork point information in the vein skeleton map corresponding to the two finger vein images to be matched;

The feature matching module 1104 is used to respectively match the branch information and fork point information of the two vein skeleton maps to obtain a plurality of matched pairs of fork points;

Fork point pair screening module 1106, used to obtain the coordinate offsets of each group of matched fork point pairs, and filter out multiple groups of target fork point pairs from each group of matched fork point pairs according to the coordinate offsets;

The length obtaining module 1108 is used to obtain the similarity of each target fork point pair, and to obtain the branch lengths of all branches in each vein skeleton diagram; the sum of similarity;

The similarity determination module 1110 is configured to obtain the similarity of the two finger vein images according to the similarity of each target fork point pair and the branch lengths of all branches, and perform identity recognition according to the similarity.

In one embodiment, the above-mentioned feature matching module 1104 is specifically configured to obtain the coordinate offset of each group of fork point pairs in the two vein skeleton maps, and filter out the coordinate offset from each group of fork point pairs that is less than the offset The multiple sets of fork point pairs of the threshold value form the first set of fork point pairs; the main direction angles of the branches corresponding to each group of fork point pairs in the first set of fork point pairs are obtained, and the distance between the corresponding branches of each group of fork point pairs is determined. Combination relationship, to obtain a plurality of target branch combinations; wherein, the main direction angle of any branch is the average value of the direction angles of each pixel in the branch; calculate the difference of the main direction angles of the branches under each target branch combination, from the first From the set of fork point pairs, filter out a plurality of sets of fork point pairs whose main direction angles of the branches under the target branch combination are less than the difference threshold to form a second set of fork point pairs; obtain each group of forks in the second set of fork point pairs For the branch linear correlation coefficient under the target branch combination in the point pair, multiple sets of fork point pairs whose linear correlation coefficient is greater than the correlation coefficient threshold are selected from the second fork point pair set as matching fork point pairs.

In one embodiment, the above-mentioned feature matching module 1104 is further configured to arrange and combine the corresponding branches of each group of fork points to obtain a plurality of branch combinations to be matched; if any branch corresponds to at least two branches to be matched combination, then calculate the difference between the main direction angles of the branches under each branch combination to be matched corresponding to the branch, and use the branch combination to be matched with the smallest difference as the target branch combination.

In one embodiment, the above-mentioned feature matching module 1104 is further configured to convert the two-dimensional coordinate vector of the branch corresponding to each group of fork point pairs into a one-dimensional coordinate vector; The correlation coefficient of the dimensional coordinate vector is used as the linear correlation coefficient of the branch under the target branch combination.

In one embodiment, the above-mentioned cross-point pair screening module 1106 is specifically configured to divide each group of matched cross-point pairs into multiple cross-point pair sets according to the coordinate offset; obtain the similarity of each group of matched fork-point pairs Calculate the sum of the similarities of the fork pairs in each fork pair set, and obtain the cumulative sum of the similarity of each fork pair set; add the sum of the similarities to the maximum fork point pair set that matches each group included in the set. Fork point pair, as the target fork point pair.

In one embodiment, the above-mentioned similarity determination module 1110 is specifically configured to add the similarities of each target fork point pair to obtain the sum of the similarities, and to add the branch lengths of all branches to obtain the sum of the branch lengths ; Calculate the ratio between the sum of the similarity and the sum of the branch lengths, multiply the ratio by 2, and use it as the similarity between the two finger vein images.

In one embodiment, the above-mentioned apparatus further comprises:

A fork-point pair identification module, used for determining unmatched fork-point pairs in each group of matched branches in each group of target fork-point pairs, as fork-point pairs to be matched;

Identify whether the fork point pair to be matched matches, and identify whether the branches under each target branch combination corresponding to the fork point pair to be matched match, if both match, add the fork point pair to be matched to the target fork point pair. in the collection.

In one embodiment, the above-mentioned fork-point pair identification module is specifically configured to calculate two pairs of unmatched branch endpoints when there are unmatched branch endpoints in each group of matched branches in the set formed by the target fork-point pair. The branch length difference of each branch; if the branch length difference is not greater than the length difference threshold, the end points of the two unmatched branches are used as the pair of fork points to be matched; if the branch length difference is greater than the length difference threshold, the unmatched branch end point A pseudo fork point is added to the corresponding longer branch, and the end point and the pseudo fork point of the shorter branch are used as the pair of fork points to be matched.

It should be noted that the finger vein identification device of the present application corresponds to the finger vein identification method of the present application, and the technical features and beneficial effects described in the embodiments of the above finger vein identification method are all applicable to finger vein identification. In the embodiment of the identification device, reference may be made to the description in the method embodiment of the present application for the specific content, which will not be repeated here, but is hereby declared.

In addition, each module in the above finger vein identification device can be implemented in whole or in part by software, hardware and combinations thereof. The above modules can be embedded in or independent of the processor in the computer device in the form of hardware, or stored in the memory in the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

In one embodiment, a computer device is provided, and the computer device may be a terminal, and its internal structure diagram may be as shown in FIG. 12 . The computer equipment includes a processor, memory, a communication interface, a display screen, and an input device connected by a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium, an internal memory. The nonvolatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the execution of the operating system and computer programs in the non-volatile storage medium. The communication interface of the computer device is used for wired or wireless communication with an external terminal, and the wireless communication can be realized by WIFI, operator network, NFC (Near Field Communication) or other technologies. The computer program implements a finger vein identification method when executed by the processor. The display screen of the computer equipment may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer equipment may be a touch layer covered on the display screen, or a button, a trackball or a touchpad set on the shell of the computer equipment , or an external keyboard, trackpad, or mouse.

Those skilled in the art can understand that the structure shown in FIG. 12 is only a block diagram of a partial structure related to the solution of the present application, and does not constitute a limitation on the computer equipment to which the solution of the present application is applied. Include more or fewer components than shown in the figures, or combine certain components, or have a different arrangement of components.

In one embodiment, a computer device is also provided, including a memory and a processor, where a computer program is stored in the memory, and the processor implements the steps in the foregoing method embodiments when the processor executes the computer program.

In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, implements the steps in the foregoing method embodiments.

Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be implemented by instructing relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage In the medium, when the computer program is executed, it may include the processes of the above-mentioned method embodiments. Wherein, any reference to memory, storage, database or other media used in the various embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (Read-Only Memory, ROM), magnetic tape, floppy disk, flash memory, or optical memory, and the like. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, the RAM may be in various forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

The technical features of the above embodiments can be combined arbitrarily. In order to make the description simple, all possible combinations of the technical features in the above embodiments are not described. However, as long as there is no contradiction in the combination of these technical features It is considered to be the range described in this specification.

The above-mentioned embodiments only represent several embodiments of the present application, and the descriptions thereof are relatively specific and detailed, but should not be construed as a limitation on the scope of the invention patent. It should be noted that, for those skilled in the art, without departing from the concept of the present application, several modifications and improvements can be made, which all belong to the protection scope of the present application. Therefore, the scope of protection of the patent of the present application shall be subject to the appended claims.

Claims

A finger vein identification method, characterized in that the method comprises:

Obtain the branch information and fork point information in the vein skeleton map corresponding to the two finger vein images to be matched;

Match the branch information and fork point information of the two vein skeleton maps respectively to obtain multiple matched pairs of fork points;

Obtain the coordinate offsets of each group of matched fork point pairs, and filter out multiple groups of target fork point pairs from each group of matched fork point pairs according to the coordinate offsets;

Obtain the similarity of each target fork point pair, and obtain the branch lengths of all branches in each of the vein skeleton diagrams; the similarity of the target fork point pair is the matched branch of each group in the target fork point pair The sum of the similarity;

According to the similarity of each target fork point pair and the branch lengths of all branches in each of the vein skeleton images, the similarity of the two finger vein images is obtained, and the identity is identified according to the similarity.
The method according to claim 1, wherein the branch information and the fork point information of the two vein skeleton diagrams are respectively matched to obtain a plurality of matched pairs of fork points, including:

Obtain the coordinate offsets of each group of fork point pairs in the two vein skeleton maps, and select multiple groups of fork point pairs whose coordinate offsets are less than the offset threshold from each group of fork point pairs to form the first fork point. pair collection;

Acquire the main direction angles of the branches corresponding to each group of fork point pairs in the first fork point pair set, and determine the combination relationship between the branches corresponding to each group of fork point pairs to obtain a plurality of target branch combinations; The main direction angle of a branch is the mean value of the direction angles of each pixel in the branch;

Calculate the difference of the main direction angles of the branches under each target branch combination, and from the first set of fork point pairs, filter out multiple groups of forks whose differences in the main direction angles of the branches under the target branch combination are less than the difference threshold point pairs to form a second set of fork point pairs;

Obtain the branch linear correlation coefficients under the target branch combination in each group of fork point pairs in the second fork point pair set, and filter out multiple groups whose linear correlation coefficients are greater than the correlation coefficient threshold from the second fork point pair set A pair of forks, as the matched pair of forks.
The method according to claim 2, wherein the determining the combination relationship between the corresponding branches of each group of fork point pairs to obtain a plurality of target branch combinations, comprising:

Arrange and combine the corresponding branches of each group of fork points to obtain a plurality of branch combinations to be matched;

If any branch corresponds to at least two to-be-matched branch combinations, the difference between the branch main direction angles under each to-be-matched branch combination corresponding to the branch is calculated, and the to-be-matched branch combination with the smallest difference is used as the target branch combination.
The method according to claim 2, wherein the acquiring the linear correlation coefficients of the lower branches of the target branch combination in each group of fork point pairs in the second fork point pair set comprises:

Convert the two-dimensional coordinate vector of the branch corresponding to each group of fork pairs into a one-dimensional coordinate vector;

The correlation coefficient of the one-dimensional coordinate vector of the lower branch of the target branch combination in each group of fork point pairs is calculated as the linear correlation coefficient of the lower branch of the target branch combination.
The method according to claim 1, wherein the acquiring coordinate offsets of each group of matched fork point pairs, and filtering out multiple groups of targets from each group of matched fork point pairs according to the coordinate offsets Fork pairs, including:

According to the coordinate offset, each group of matched fork-point pairs is divided into a plurality of fork-point pair sets;

Obtain the similarity of each group of matched fork-point pairs, calculate the sum of the similarities of the fork-point pairs in each fork-point pair set, and obtain the cumulative sum of the similarity of each fork-point pair set;

Each group of matched fork-point pairs included in the fork-point pair set with the maximum similarity sum is taken as the target fork-point pair.
The method according to claim 1, wherein the similarity of the two finger vein images is obtained according to the similarity of each target fork point pair and the branch lengths of all branches in each of the vein skeleton images degrees, including:

The similarity of each of the target fork pairs is added to obtain the sum of the similarity, and the branch lengths of all branches are added to obtain the sum of the branch lengths;

Calculate the ratio between the sum of the similarity and the sum of the branch lengths, and multiply the ratio by 2 to obtain a value as the similarity of the two finger vein images.
The method according to claim 1, wherein after filtering out multiple groups of target fork point pairs from each group of matched fork point pairs according to the coordinate offset, the method further comprises:

Determine the unmatched fork pairs in the matched branches in each group of target fork pairs as the fork pairs to be matched;

Identifying whether the pair of fork points to be matched matches, and identifying whether the branches under each target branch combination corresponding to the pair of fork points to be matched match, if both match, adding the pair of fork points to be matched in the set formed by the target fork pair.
The method according to claim 7, wherein the determining unmatched branches in each group of target fork point pairs, and acquiring unmatched fork point pairs in the unmatched branches as fork point pairs to be matched ,include:

When there is an unmatched branch end point in each group of matched branches in the set formed by the target fork point pair, calculate the branch length difference between the two branches corresponding to the unmatched branch end point;

If the branch length difference is not greater than the length difference threshold, the two unmatched branch end points are used as the pair of fork points to be matched;

If the branch length difference is greater than the length difference threshold, a pseudo fork point is added to the longer branch corresponding to the end point of the unmatched branch, and the end point of the shorter branch and the pseudo fork point are used as the waiting point. Match fork pairs.
A finger vein identification device, characterized in that the device comprises:

a feature acquisition module, used for acquiring branch information and fork point information in the vein skeleton map corresponding to the two finger vein images to be matched;

The feature matching module is used to match the branch information and fork point information of the two vein skeleton maps respectively to obtain multiple sets of matched fork point pairs;

A cross-point pair screening module, used to obtain the coordinate offsets of the matched cross-point pairs of each group, and screen out multiple groups of target cross-point pairs from the matched cross-point pairs according to the coordinate offsets;

a length obtaining module, used to obtain the similarity of each target fork point pair, and obtain the branch lengths of all branches in each of the vein skeleton diagrams; the similarity of the target fork point pair is the target fork point pair The sum of the similarity of the matched branches in each group;

The similarity determination module is configured to obtain the similarity of the two finger vein images according to the similarity of each target fork point pair and the branch lengths of all the branches, and perform identity recognition according to the similarity.
A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, when the processor executes the computer program, the steps of the method according to any one of claims 1 to 8 are implemented.
A computer-readable storage medium on which a computer program is stored, characterized in that, when the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.