KR20040105065A - Strong correlated or weakly correlated grouping fingerprint matching method - Google Patents

Abstract

PURPOSE: A method for matching fingerprints through grouping of minutiae having strong and weak relationship is provided to match the fingerprints with triangles by grouping the minutiae having the strong relationship and generating the triangles formed with 3 minutiae of each group. CONSTITUTION: When the fingerprints are matched, feature data is extracted. The triangles are generated by combining 3 minutiae for each cluster of the registered and the inputted fingerprint(701). The matched triangles are extracted by matching the triangles of both fingerprints and the similar clusters are extracted based on the extracted triangles(703). For both fingerprints, triangle pairs made to 2 combinations are generated to each cluster from the extracted triangles(707). The corresponding triangle pair is matched from the triangle pairs of both fingerprints(709). A similarity is calculated based on a matching result(711). If the similarity is over a predetermined value, the matching is judged as the successful matching(715).

Description

Fingerprint matching method through strong and weak correlation grouping {STRONG CORRELATED OR WEAKLY CORRELATED GROUPING FINGERPRINT MATCHING METHOD}

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a fingerprint matching method, and more particularly to a method for generating a cluster with feature points existing on the same ridge or neighboring ridges, and for performing fingerprint matching (or matching) with triangles generated for each cluster. will be.

In recent years, the number of password theft and abuse over the Internet is increasing. As such, there is an increasing demand to replace personal identification and authentication with biometrics in order to prevent criminal acts that cause personal property disadvantages through electronic commerce.

In general, biometrics (Biometrics) refers to a technology that uses the unique characteristics of each person as a specific unit for identification based on human physiological or behavioral characteristics. In other words, it is a technical field that uses human biological characteristics for identification through an automated device. Biometric technology is safer and more convenient than identity verification using existing passwords or passwords because it requires the person to be recognized at the time of recognition and eliminates the need to remember passwords or carry tokens. The bio-parts used for biometrics are diverse such as fingerprint, retina, iris, face, hand, vein, voice, signature, body smell, DNA, etc. However, the most popular parts are bio-prints using fingerprint, voice, iris, and face. It is awareness.

In particular, fingerprint recognition is one of the oldest technologies that has been used successfully in many applications. Fingerprints are formed by the uplift of the sweat glands to form a constant flow. Retina and iris in reliability and stability of identification performance due to their unique characteristics that are different from person to person and do not change for life. It is estimated to be higher than the vein, facial, and the like, and is used as an efficient personal authentication method.

In the conventional fingerprint recognition method, the feature-based fingerprint recognition is mainly used for comparative evaluation using the feature points. Here, the feature points refer to the end points of the ridges (that is, the point where the ridges start and end) and the branching points of the ridges, which are used as data for fingerprint matching. Among the matching methods, there is a method of evaluating a matching score through rotation and movement through a map of ridges and a map of feature points. However, the statistical matching method for such a map has a disadvantage in that it is relatively slow in time because of a large amount of data. Therefore, a method of reducing the amount of data and quickly performing matching by extracting and comparing geometric position and direction information using only feature points is used. However, methods using only feature points still have problems.

For example, among the methods using the feature points on the same ridge, there is a method of identifying by the distance between the two feature points and the center point scattering. However, when the ridge is damaged and cut, the same ridge information is lost and the error occurs more severely. The more severe the change of direction and the smaller the interval, the larger the margin of error. Therefore, when using the method using the feature points has been pointed out that the matching performance is poor when the number of feature points is small. If the feature points are small, the image size of the input fingerprint is small or the number of feature points is small and the fingerprints are damaged. There is a problem in that an error is generated and an appropriate recognition rate cannot be guaranteed.

In order to solve the problems described above, an object of the present invention is to group feature points having a strong correlation, generate triangles consisting of three feature points for each group, and perform fingerprint matching with the triangles. It provides a method for doing so.

Another object of the present invention is to provide a method for generating a cluster with feature points existing on the same ridge, generating triangles composed of three feature points for each cluster, and performing fingerprint matching with the triangles.

According to the present invention for achieving the above object, the fingerprint recognition matching method through the correlation grouping, the process of checking the number of feature points present on the ridge for each ridge at the time of fingerprint registration, the number of the feature points is at least In case of four, a process of generating a strong correlation cluster with feature points existing on the ridge, and extracting four feature points not included in the cluster generation from neighboring ridges from a core center, creates a strong correlation cluster. And generating a weak correlation cluster by collecting branch points among the feature points not included in the cluster generation, and generating another weak correlation cluster by collecting end points of the feature points. To store the information of the created clusters as feature data of the registration fingerprint Extracting the feature data from an input fingerprint, generating triangles by combining three feature points per cluster for each of the registration fingerprint and the input fingerprint, and triangles of the registration fingerprint. And extracting matching triangles by matching triangles of the input fingerprint, extracting similar clusters based on the extracted triangles, and extracting two triangles from the extracted triangles for each of the registration fingerprint and the input fingerprint. Generating triangle pairs that can be combined by cluster, matching triangle pairs corresponding to each other among the triangle pairs of the generated fingerprint and the triangle pair of the input fingerprint, and calculating similarity based on the matching result. The process of calculating and the similarity is said to be a matching success when more than a predetermined reference value It characterized by including the step of.

Preferably, the information of the cluster, characterized in that it comprises the coordinates, types, directions and piers of the feature points present on the same ridge.

Preferably, the triangle extraction process, for each of the registration fingerprint and the input fingerprint, the process of obtaining the angles, the directions of the vertices, the bridges of the vertices of each triangle, and the triangles and input fingerprint of the registration fingerprint Comparing the triangles of the triangles, the directions of the vertices, the intersections of the vertices, characterized in that it comprises the step of extracting the triangles matching the predetermined error range.

Advantageously, the similarity calculation process calculates a distance between the centers of two triangles forming a triangle pair for each of the registration fingerprint and the input fingerprint and the angle between the centers and two straight lines connecting the center core. Comparing the triangle pairs of the enrolled fingerprint with the triangle pairs of the input fingerprint and checking whether the distance and the angle correspond to each other within a predetermined error range, and increasing the matching score if the match is within the predetermined error range. And calculating the similarity based on the matching score.

Preferably, the matching score is increased to different specific gravity values according to the type of cluster to be compared.

1 is a block diagram of a fingerprint recognition device according to an embodiment of the present invention.

2 is a view for explaining a general fingerprint image.

3 is a view for explaining a cluster generation method according to the present invention.

4 is a diagram illustrating a cluster creation procedure according to an embodiment of the present invention.

5 is a diagram for explaining a method of obtaining a bridge;

6 is a diagram showing a structure of a database for storing feature data consisting of information of a plurality of clusters.

7 is a diagram illustrating a fingerprint matching procedure according to an embodiment of the present invention.

8 shows triangles that can be made up of three feature points for a cluster;

9 is a view for explaining the fingerprint matching in accordance with the present invention.

<Explanation of symbols for the main parts of the drawings>

10: fingerprint image input device 20: A / D converter

30: Image memory 40: I / O interface

50: console 60: external storage device

70: main controller 80: working memory

90: program memory 100: output device

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. In describing the present invention, when it is determined that the detailed description of the related known function or configuration may unnecessarily obscure the subject matter of the present invention, the detailed description thereof will be omitted.

Hereinafter, the present invention will be described a technique for generating strong correlation clusters and weak correlation clusters, and performing fingerprint matching with triangles generated for each cluster.

1 is a block diagram of a fingerprint recognition device according to an embodiment of the present invention.

As shown, the fingerprint recognition system, the A / D converter 20, the image memory 30, the I / O interface 40, the console 50, the external storage device 60, the main control unit 70 , A working memory 80, a program memory 90, and an output device 100. Here, reference numeral 10 denotes a heterogeneous fingerprint input sensor (different types of fingerprint input sensors) that can be coupled to the fingerprint recognition system.

Referring to FIG. 1, first, the fingerprint input sensor 10 scans a fingerprint of a finger contacting an upper surface of a panel and outputs an analog signal according to the contrast of the fingerprint image to the A / D converter 20. The A / D converter 20 converts an analog signal applied from the fingerprint image input apparatus 10 into a digital signal and outputs the digital signal to the image memory 30. That is, the A / D converter 20 converts an analog signal corresponding to each pixel (or pixel) into a bit string having a predetermined number of bits and outputs the converted bit signal. The image memory 30 stores digital data about a fingerprint image applied from the fingerprint input sensor 10. The I / O interface 40 matches the image memory 30 with the console 50, the output device 100, and the main controller 70.

The console 50 includes a plurality of operation keys, and outputs a key signal according to a user's operation to the main controller 70. The external memory device 60 stores a plurality of input fingerprint image data (characteristic data (information of clusters)) of a registered fingerprint. The main control unit 70 performs an operation mode according to the key signal input from the console 50. When the operation mode is the authentication mode, the main control unit 70 displays the registered local culture image (characteristic data) according to identification information (eg, ID, employee number, resident number, etc.) input by the user through the console 50. It reads from the memory device 60, compares the read registration paper image with the input paper culture image currently input from the fingerprint input sensor 10, and outputs the result through the output device 100.

The work memory 80 is a memory used when comparing a currently input fingerprint image with a pre-registered fingerprint image. The program memory 90 stores program data and initial data for controlling the overall operation of the fingerprint recognition system. In particular, an algorithm for extracting and matching (matching) features of a fingerprint according to the invention is programmed. The output device 100 displays an overall state of the fingerprint recognition system and input user input information and the like on a display window (eg, a liquid crystal display (LCD)) under the control of the main controller 70. In particular, according to the present invention, a message or a corresponding fingerprint image according to the comparison result of the input fingerprint image and the comparison target fingerprint image is displayed.

The operation of the fingerprint recognition system configured as described above will be described in detail with reference to the accompanying drawings.

2 is a view for explaining a general fingerprint image.

As shown, the fingerprint image is divided into the ridge 100, the valley 110, the core 140 that is the fingerprint center, and the feature points present on the ridge 110. As described above, the feature point is divided into the ridge end point 120 and the branch point 130. The present invention defines a cluster of at least four feature points existing on each same ridge or neighboring ridge as a cluster, and the data of clusters (quantity of feature points, coordinates of specific points, type, direction, piers, etc.) Are stored as feature data.

3 is a view for explaining a cluster extraction method according to the present invention.

As shown, all feature points are extracted and labeling 300, 310 is performed for each ridge. Then, a set of at least four feature points existing on the same ridge or neighboring ridges is defined as a cluster, and a sequence number is assigned to each cluster. The clusters are numbered in descending order according to the number of feature points. If there is a feature point that does not form a cluster (when there are few feature points of the fingerprint and the number of feature points due to damage), the feature points corresponding to the ridge end points are collected to form an end point cluster 320 and the branch point of the ridges. The corresponding feature points are collected to form the branch cluster 330. As such, the present invention forms a cluster in consideration of the characteristics of the feature points.

4 illustrates a cluster creation procedure according to an embodiment of the present invention.

First, the main control unit 70 stores the fingerprint image input from the fingerprint input sensor 10 in the work memory 80. Here, the analog fingerprint image from the fingerprint input sensor 10 is converted into a digital fingerprint image of a predetermined format in the A / D converter 20 and stored in the image memory 30, and the digital fingerprint stored in the image memory 30. The image is loaded into the working memory 80 under the control of the main controller 70. The fingerprint image stored in the work memory 80 is converted into a fingerprint image in which ridges are expressed in units of 1 pixel through various ring processes (normalization, segmentation, filtering, binarization, sessionization, etc.). Thereafter, the main controller 70 generates and stores a cluster from a fingerprint image in which the ridge is expressed in units of 1 pixel. Looking at this in detail as follows.

Referring to FIG. 4, in operation 401, the main controller 70 finds a central core 340 in the fingerprint image. In operation 403, the main control unit 70 performs labeling on the ridges in the order from the central core 340. In operation 405, the main control unit 70 extracts minutia from the fingerprint image. As mentioned above, the feature point may be a bifurcation at which the ridge splits and an end point (or a disadvantage) at which the ridge breaks. In this process, the main control unit 70 obtains the coordinates (x, y), the type (type), the direction (θ), the piers for each feature point and stores it as feature point information. The pier acquisition method will be described in detail later with reference to FIG. 5.

In operation 407, the main controller 70 creates a feature point list for a specific ridge (the ridge in the corresponding order). That is, a list of feature points existing on the ridge is stored corresponding to the ridge. In operation 409, the main controller 70 searches for the feature point list of the ridge to check whether the number of feature points present on the ridge is greater than or equal to "4".

If greater than or equal to "4" (at least 4 or more), the main control unit 70 proceeds to step 411 to create a ridge cluster (or a strong correlation cluster), and then returns to step 407 to the ridge of the next order. Create a list of feature points for. Here, the ridge clusters will be described in detail later with reference to FIG. 6. On the other hand, if the number of feature points present on the ridge is less than "4", the main control unit 70 proceeds to step 413 to check whether the cluster creation for the last ridge is completed. If the cluster preparation for the last ridge is not completed, the main control unit 70 returns to step 407 to create a feature point list for the next ridge.

If the cluster for the last ridge is completed, the main controller 70 proceeds to step 415 to collect the branch points among the feature points not included in the cluster formation, and in step 417 the branch point cluster (or the branch point cluster with the branch points). Create a weak correlation cluster. The branch cluster will be described in detail later with reference to FIG. 6. After creating the branch point cluster, the main control unit 70 collects endpoints of the feature points not included in the cluster formation in step 419, and ends with the collected endpoints in step 421 (or weak correlation cluster). Write. The endpoint cluster will be described in detail later with reference to FIG. 6.

In operation 423, the main controller 70 arranges the clusters (including the ridge cluster, the branch point cluster, and the end point cluster) previously generated according to the number of feature points. For example, order numbers in descending order. In operation 425, the main controller 70 ends after storing the sorted clusters in a database. As such, the set of clusters stored in the database is called feature data.

In addition, four feature points are extracted from neighboring ridges from the core center to create one strong correlation cluster.

5 is a view for explaining the pier acquisition method.

Referring to FIG. 5, first, an equation 520 of a straight line between the center core 530, the branch feature point 550, and the endpoint feature point 500 is obtained. The equation of the tangent line is obtained through curve fitting of the ridge at the branch feature point 550 and the endpoint feature point 500. Thereafter, the angular angle 560 is calculated using the slope m_1 of the ridge at the feature point and the slope m_2 of the linear anticorrosion type 520 formed by the central core 530 and the feature points 500 and 550. The curve fitting equation at the feature point present on the ridge is calculated using a quadratic equation as shown in Equation 1 below.

The determinant for obtaining the coefficient in Equation 1 is as shown in Equation 2 below.

And the coefficient of a quadratic equation is calculated | required as <Equation 3>.

On the other hand, the equation 570 of the tangent of the feature points (500, 550) is as shown in Equation (4).

Finally, the piers are calculated as shown in Equation 5 with the slope of the tangent line m_1 with respect to the feature point and the slope of the straight line between the center core (m_2).

6 shows a structure of a database for storing feature data consisting of information of a plurality of clusters.

As shown, the database includes a first area 600 in which the quantity of clusters is largely stored, and second areas 610-630 in which information of clusters are stored in order. The second area includes a plurality of cluster areas for storing information of a plurality of clusters (ridge cluster, branch cluster, and endpoint cluster). Each cluster area stores an area 611 in which a cluster number is stored, an area 612 in which a feature point quantity is stored, an area 613 and 614 in which feature point coordinates are stored, an area 615 in which a feature point type is stored, and a specific load direction. And an area 517 in which the feature point pier is stored. Here, the regions 613-617 that store the feature point coordinates, types, directions, and piers are configured in plural numbers according to the number of feature points.

A method of performing fingerprint matching based on a database (or feature data) stored as described above is as follows.

7 illustrates a fingerprint matching procedure according to an embodiment of the present invention.

Referring to FIG. 7, first, in operation 701, the main controller 70 generates triangles that can be combined with three feature points for each cluster for the input fingerprint and the registration fingerprint, and stores the generated triangle information. Here, the number m of triangles that can be generated when the number of feature points included in a specific cluster is n is calculated as in Equation 6 below, and this is shown in the accompanying drawings as shown in FIG. 8.

In operation 703, the main controller 70 extracts similar clusters by comparing the triangles of the registration fingerprint and the triangles of the input fingerprint. First we extract triangles whose triangles, vertices directions and piers coincide. If a predetermined ratio of all the triangles in the cluster coincide with the predetermined ratio, the similar cluster is determined. The main control unit 70 proceeds to step 705 to generate triangle pairs that can be made in two combinations from the extracted triangles of the registered fingerprint for each cluster, and to proceed to step 707 to triangles of the extracted input fingerprint. Generates triangle pairs in clusters that can be made in two combinations.

In operation 709, the main control unit 70 matches a triangle pair of an input fingerprint and a triangle pair of a registration fingerprint. That is, as illustrated in FIG. 9, the first distance 720 connecting the centers of the two triangles of the input fingerprint is calculated, and the angle between the two straight lines connecting the centers and the core center 740 is defined. 730). Similarly, calculate a second distance 750 connecting the centers of the two triangles of the registration fingerprint, and calculate the angle between the two straight lines connecting the centers and the core center 790 (second angle, 780). do. Thereafter, the first distance and the second distance, and the angle between the first and second angles are all checked within an error range. If the match is within the error range, the main control unit 70 increases the matching score. Here, the increment has a different value depending on the type of cluster. That is, a greater weight is placed on the strong correlation cluster than on the weak correlation cluster.

In operation 711, the main controller 70 calculates a similarity between the input fingerprint and the registered fingerprint by a predetermined method based on the matching score. In operation 713, the main controller 70 checks whether the calculated similarity is greater than a predetermined reference value. If the similarity is greater than the reference value, the main control unit 70 proceeds to step 715 and displays the matching success message on the display window and then terminates. On the other hand, when the similarity is less than or equal to the reference value, the main controller 70 proceeds to step 717 and ends displaying the matching failure message. Thereafter, the main controller 70 transfers the matching result to the host, for example, to allow or deny the user access to the system.

Meanwhile, in the detailed description of the present invention, specific embodiments have been described, but various modifications may be made without departing from the scope of the present invention. Therefore, the scope of the present invention should not be limited to the described embodiments, but should be defined not only by the scope of the following claims, but also by those equivalent to the scope of the claims.

As described above, the fingerprint matching method according to the present invention reduces errors by grouping correlated feature points and different weight values for matching scores when comparing strong and weak correlation clusters in fingerprint matching. Can be. In the case of others, the feature points form different clusters, and comparing clustering versus clustering is more effective in reducing the matching error because the weaker correlation cluster is involved in the matching than the cluster with the strong correlation. In the case of the same person, when the clusters are compared with each other, a strong correlation cluster is mainly involved in the matching, which can effectively increase the accuracy of the self-determination. Characteristic points of the individual can identify the fingerprint. In addition, the number of matching cases increases due to the increase in the number of feature points due to the damage of the fingerprints (wounds, cracks, abrasions, etc.), thereby increasing the accuracy of fingerprint identification by lowering the false recognition rate by the clustering method.

Claims (6)

In the fingerprint recognition matching method through correlation grouping, At the time of fingerprint registration, checking the number of feature points on the ridge for each ridge; When the number of feature points is at least four, creating a strong correlation cluster with the feature points present on the ridge; Generating a strong correlation cluster by extracting four feature points not included in the cluster generation from neighboring ridges from a core center; Generating a weak correlation cluster by collecting branch points among the feature points not included in the cluster generation, and generating another weak correlation cluster by collecting endpoints of the feature points; Storing the generated cluster information as feature data of a registration fingerprint; During fingerprint matching, extracting the feature data from an input fingerprint; Generating triangles by combining three feature points per cluster for each of the registration fingerprint and the input fingerprint; Extracting matching triangles by matching triangles of the registration fingerprint and triangles of the input fingerprint, and extracting similar clusters based on the extracted triangles; Generating triangle pairs for each of the registration fingerprint and the input fingerprint for each of the triangles, which can be formed in two combinations, for each cluster; Matching triangle pairs corresponding to each other among the triangle pairs of the generated fingerprint and the triangle pair of the input fingerprint, and calculating a similarity based on the matching result; And determining that the similarity is a matching success at a time equal to or greater than a predetermined reference value. The method of claim 1, And the cluster information includes coordinates, types, directions, and piers of feature points. The method of claim 2, The piers are calculated as shown in Equation 7 with the slope m_1 of the tangent line to the corresponding feature point and the slope m_2 of the straight line connecting the feature point and the center core. The method of claim 1, wherein the triangle extraction process, Obtaining angles of the respective triangles, directions of the vertices, and piers of the vertices for each of the registration fingerprint and the input fingerprint; And comparing triangles of the enrolled fingerprint with triangles of an input fingerprint and extracting triangles whose angles, vertices, and corners of the vertices coincide within a predetermined error range. The method of claim 1, wherein the similarity calculation process, Calculating a distance between centers of two triangles forming a pair of triangles for each of the registration fingerprint and an input fingerprint, and an angle between two straight lines connecting the centers and the central core; Comparing the triangle pairs of the registration fingerprint with the triangle pairs of the input fingerprint and checking whether the distance and the angle correspond to each other within a predetermined error range; Increasing the matching score if the match is within the predetermined error range; And calculating the similarity based on the matching score. The method of claim 5, The matching score is increased by different specific gravity values according to the type of cluster to be compared.