KR100397916B1 - Fingerprint registration and authentication method - Google Patents

Abstract

The present invention relates to a fingerprint registration and authentication method, and more particularly, when a fingerprint is registered in a fingerprint recognition system using a feature point of a fingerprint, a fingerprint is input a plurality of times, and the feature points of the fingerprint that are input each time are synthesized and reliability of the feature points is obtained. The present invention relates to a fingerprint registration and authentication method for improving the speed and accuracy of a fingerprint recognition system by evaluating and matching the location information and the direction information of the synthesized feature points using the reliability of each feature point. In the fingerprint authentication method according to the present invention, by receiving an image of a fingerprint to be registered N times, extracting a feature point for each of the N fingerprint images received, calculating the position coordinates of the feature points and the direction of the ridge in the position coordinates. Generating a; Synthesizing the N feature data and giving reliability to each feature point in the synthesized feature data; And registering the coordinates of the synthesized feature points, the direction of the fingerprint ridge at the location, and the reliability of the feature points to generate the feature data, and then storing the feature data in the fingerprint data storage means.

Description

Fingerprint registration and authentication method

The present invention relates to a fingerprint registration and authentication method, and more particularly, when a fingerprint is registered in a fingerprint recognition system using a feature point of a fingerprint, a fingerprint is input a plurality of times, and the feature points of the fingerprint that are input each time are synthesized and reliability of the feature points is obtained. The present invention relates to a fingerprint registration and authentication method for improving the speed and accuracy of a fingerprint recognition system by evaluating and matching the location information and the direction information of the synthesized feature points using the reliability of each feature point.

One of the security authentication methods using biometric information is fingerprint recognition. Fingerprint recognition technology is a technology that determines whether or not a registered person by recognizing a fingerprint showing a unique characteristic difference for each person by an image processing method. The fingerprint recognition technology is divided into a "registration process" for extracting a feature of a fingerprint and storing it as feature data, and a "matching process" for extracting a feature of an authentication target fingerprint and comparing it with the stored feature data.

Figure 1 shows an overview of the fingerprint recognition technology, (a) fingerprint registration process, (b) shows the processing flow of the matching process.

First, the fingerprint registration process (Fig. 1 (a)) starts by reading a fingerprint by the fingerprint input device 1 to obtain a fingerprint image. There are two types of fingerprint input devices: optical using an optical scanner and non-optical using a semiconductor sensor. CCD and CMOS are used as image acquisition sensors.

The fingerprint image acquired by the fingerprint input device 1 is subjected to processing such as removing the miscellaneous images included in the fingerprint image in the preprocessing step 2 and thinning and highlighting the feature points. To facilitate.

The feature extraction step (3) is a process of extracting feature points that can distinguish each fingerprint. The feature extraction step (3) identifies the position, direction, and shape of the end point (end point) of the ridge and the branch point (described through Figs. 2A and 2B). Extracted as feature data (fingerprint data). The fingerprint data extracted in this way is stored in the fingerprint data DB 4.

Meanwhile, the matching process shown in FIG. 1B will be described. When a person who wants to obtain fingerprint authentication acquires a fingerprint image from the fingerprint input device 1 ', the acquired fingerprint image is preprocessed (2'), and feature points are extracted (3 ') and stored in the fingerprint data DB 4'. After comparing the original fingerprint data with the feature 5, it is determined whether the newly acquired fingerprint image is consistent with the registered fingerprint data (6) and a matching process of fingerprint authentication is performed.

Feature extraction in the fingerprint registration and matching process as shown in FIG. 1 will be described in more detail. As the characteristic point of the fingerprint, the branch point 21 and the disadvantage 22 included in the fingerprint are used as shown in the circled portion of FIG. 2A and the enlarged view of FIG. 2B.

In the fingerprint registration and matching process, the humidity of the fingerprint, the pressure of the finger, the position of the finger in contact with the fingerprint input window, and the like may be changed at every fingerprint input, and thus the fingerprint image acquired will be different. Very often things happen that may or may not be extracted. Therefore, when fingerprints are registered using these feature points, feature data is generated by repeatedly extracting feature points from one fingerprint two or three times as shown in FIG. That is, as shown in FIG. 3A, the positions of the feature points 31 in the feature data 32 extracted once, the feature data 33 extracted twice, and the feature data 34 extracted three times. Even if is different from each other by matching all the feature points to minimize the misrecognition due to the fingerprint state failure at the time of fingerprint registration.

The information about the position of each feature point 31 of the feature data 32, 33, and 34 extracted three times and the direction of the fingerprint ridge at the location are stored in the fingerprint template DB 35 as fingerprint data, respectively. It is registered. 3, three fingerprint data are stored under a specific ID.

Next, the matching process will be described. When the feature point is extracted from the input fingerprint inputted by the person who wants to be authenticated, as shown on the left side of FIG. 3 (b), the feature data 37 having the position of the feature point 36 of the input fingerprint and the direction of the ridge at that position is shown. ) Is extracted. Matching is performed on the feature points 33 of the feature data 37 of these input fingerprints and the feature points 31 of the three feature data 32 ', 33', 34 'stored in the fingerprint template DB 32, respectively. The score is calculated according to the number of matching feature points.

That is, as shown in Fig. 3B, the feature points matched between the feature data 32 'registered once and the input fingerprint feature data 37 are displayed in a circle, and the feature data 33' registered twice. ) And the feature points matched between the input fingerprint feature data 37 are indicated by a rectangle, and the feature points matched between the feature data 34 'registered three times and the input fingerprint feature data 37 are indicated by a triangle. . If the scores are scored for the number of matched feature points 31, 36 and at least one of each feature data 32 ', 33', 34 'is greater than a predefined threshold, the input fingerprint is assigned to the registered fingerprint. It is determined by the same fingerprint.

In the above-described conventional fingerprint registration and matching method, the more fingerprints are registered, the better the fingerprint recognition performance. However, as the number of registrations increases, the number of times that a match must be increased increases the time required for fingerprint authentication. In addition, the feature points extracted from the input fingerprint may be noise due to the fingerprint state at the time of input. In the conventional method, matching is performed without considering the possibility of extracting the feature point due to such noise. have. (According to the method described above, even noise is treated as if it were a normal feature point).

In order to solve the problem of the conventional fingerprint registration and matching, the fingerprint recognition system using the feature of the fingerprint, when registering a fingerprint a plurality of times to receive a fingerprint multiple times each time to synthesize the feature points of the input fingerprint and improve the reliability of the feature points The purpose of the present invention is to provide a fingerprint registration and authentication method that increases the speed and accuracy of a fingerprint recognition system by matching and matching the position coordinates and direction information of the synthesized feature points using the reliability of each feature point.

1 is a flow chart showing a general fingerprint authentication method

2A and 2B show feature points of a fingerprint

Figure 3 is a flow chart showing one embodiment of a conventional fingerprint authentication method

4 is a flowchart showing a fingerprint authentication method of the present invention.

<Description of Drawing>

31 36 41 48: feature point, 21: branch point, 22: disadvantage, 45: synthesized feature data

<Overview of invention>

1. Registration process

Also in the fingerprint registration process according to the present invention, the image (fingerprint image) of the fingerprint to be registered is received a plurality of times (N times) as in the prior art. Feature data are extracted for each of the N fingerprint images received, and the feature data is generated by calculating the position coordinates of the feature points and the direction of the ridges in the position coordinates.

Then, N feature data are synthesized. Then, feature points of each feature data are overlapped, and there may be feature points included in each feature data, or feature points included in only one feature data. Of course, ideally all feature points should be included in every feature data entered each time, but in practice this is not the case.

Next, the reliability of each feature point is calculated from the feature data in which the N feature data are synthesized. For example, the reliability of each feature point may be calculated by giving the highest weight to a feature point belonging to all N feature data and giving the lowest weight to a feature point belonging to only one feature data.

As described above, the registration process is performed by generating the feature data by templateting the position coordinates of the synthesized feature points, the direction of the fingerprint ridge at the position and the reliability of the feature points, and storing the feature data in the fingerprint data storage means.

2. Matching Process

The concept of comparing and matching feature data (one feature data having N feature data synthesized) stored in the fingerprint data storage means with the feature data of a new fingerprint input is the same as in the conventional method. However, in the prior art, N feature data stored in the fingerprint data storage means and feature data of the input fingerprint had to be matched N times. However, in the present invention, only one feature data stored in the fingerprint data storage means is matched with the input fingerprint. Just do it.

Since the feature data stored in the fingerprint data storage means includes the reliability data of each feature point, it is possible to calculate the reliability of the matched feature points between the feature data of the input fingerprint and the registered feature data at the time of matching, If the reliability is higher than the predetermined reference value, it is determined that the matching is successful and fingerprint authentication can be performed. That is, the matching process in the present invention basically compares feature data stored in the fingerprint template DB with feature data of a newly input fingerprint, finds feature points that are matched (matched), and calculates how reliable the feature points are. By doing so, it is determined whether the input fingerprint is authenticated.

<Example>

Hereinafter, an embodiment in which the technical idea of the present invention is embodied will be described with reference to the drawings (Fig. 4). Fig. 4 is a flowchart showing an embodiment of the fingerprint registration and matching method according to the present invention, showing an embodiment in which the number of fingerprint inputs N is 3;

First, the registration process shown in FIG. 4A will be described.

First, the image of the fingerprint to be registered is input three times, and the feature points 41 are extracted for each of the three fingerprint images, and the position coordinates of each feature point 41 and the direction of the ridge in the position coordinates are calculated. Generate data 42, 43, 44.

Next, three feature data 42, 43, and 44 are synthesized. The synthesized feature data 45 finds feature points commonly included in feature data of two or three fingerprint images. In FIG. 4, the feature points indicated by circles represent feature points commonly belonging to any two feature data, and the feature points indicated by triangles represent feature points belonging to all three feature data.

The reliability of the feature points of the synthesized feature data 45 is given. There may be a variety of methods for assigning reliability, but a weighting method may be used according to the overlapping degree of each feature point. For example, in the composite feature data 45 of FIG. 4, the feature points indicated by the triangles belong to all three feature data, and thus, "3" is assigned as the weight and the feature points indicated by the circle are assigned the weight "2". The remaining feature points may be given a weight "1" (strictly not a weight).

Then, the position coordinates of the feature points of the synthesized feature data 45, the direction of the fingerprint ridge at the position and the weight data applied to the feature points are templated and stored in the fingerprint template DB 46. 4, it can be seen that unlike the conventional method, one feature data is stored in the fingerprint template DB 46 under a single ID.

Next, the matching process shown in Fig. 4B will be described. The input fingerprint feature data 47 from which the newly input feature points 48 are extracted is compared with the synthesized feature data 49 retrieved from the fingerprint template DB 46. Compare and add the weights of the matched feature points. That is, 2 × 3 = 6 is calculated since the weight of the feature point indicated by the triangle in FIG. 4 is 2 and two matches, and 4 × 2 = 8 is calculated because the weight of the feature point indicated by the circle is 2 and it is matched 4 . The remaining five feature points have a weight of 1, so 5 × 1 = 5. Therefore, the total reliability of the feature points calculated in the embodiment of FIG. 4 is 6 + 8 + 5 = 19. If the reference reliability value required for fingerprint authentication is set to 15 in advance, the fingerprint authentication is normally performed.

In the above, embodiments of embodying the technical idea of the present invention have been described. It is a matter of course that the technical idea of the present invention is not limited to the above-described embodiment. The technical scope of the present invention is determined by the appended claims.

When the fingerprint is recognized, the state of the input fingerprint may be largely varied, and thus the recognition performance of the fingerprint recognition system varies greatly according to the amount of information that the registered fingerprint feature points have. According to the fingerprint registration and matching method of the present invention, when a fingerprint is registered several times, the recognition rate is improved as the fingerprint is registered several times, and there is an advantage in that only one matching is performed without multiple matching.

In addition, according to the present invention, by giving reliability information to each feature point according to the frequency appearing when registering a fingerprint multiple times, matching can be performed considering the possibility of being a normal feature point or noise, thereby greatly improving the recognition rate. That is, in the case of noise, since the probability of overlap for all feature data is almost zero, reliability is lower than other normal feature points in terms of reliability, and thus the degree of contribution to the reliability score is very small in the matching process.

Claims (6)

Fingerprint authentication consisting of a registration process of extracting a feature of a fingerprint and storing and registering it as feature data in a fingerprint data storage means, and a matching process of determining whether fingerprint authentication is performed by extracting a feature of an authentication target fingerprint and comparing it with registered feature data. As a method, the registration process Receiving the image of the fingerprint to be registered N times, extracting the feature points for each of the N fingerprint images received, calculating the position coordinates of the feature points and the direction of the ridge in the position coordinates to generate the feature data; Synthesizing the N feature data and giving reliability to each feature point in the synthesized feature data; And generating the feature data by templateting the position coordinates of the synthesized feature points, the direction of the fingerprint ridge at the location, and the reliability of the feature points, and storing the feature data in the fingerprint data storage means. The method of claim 1, wherein in assigning reliability to the feature points, Give the highest N-th weight to the feature points belonging to all N feature data, and give the lower (N-1) weights to feature points belonging to (N-1) feature data, any one feature Fingerprint registration and authentication method characterized by giving the lowest first weight to the feature points belonging to only data. The method of claim 2, wherein the matching process Comparing feature data (registration fingerprint data) stored in the fingerprint data storage means with feature data (input fingerprint data) of a new fingerprint input; Calculating reliability according to the number of matched feature points as a result of comparing the registered fingerprint data and the input fingerprint data; And determining whether fingerprint authentication is performed if the calculated reliability is higher than a predetermined reference value. The method of claim 3, wherein in calculating the reliability according to the number of matched feature points, the reliability is The number of feature points of the input fingerprint data that matches all the feature points belonging to the N feature data of the registered fingerprint data multiplied by the N-th weight, and the feature points commonly belonging to the (N-1) feature data of the registered fingerprint data. The first weight value is multiplied by the number of feature points of the input fingerprint data matched with the (N-1) weighted value, and the number of feature points of the input fingerprint data matched with the feature point belonging to one feature data of the registered fingerprint data. Fingerprint registration and authentication method characterized in that the sum of the multiplied value is calculated. The fingerprint registration and authentication method according to any one of claims 1 to 4, wherein N = 3. The fingerprint registration and authentication method according to claim 2 or 4, wherein the Nth weight is 3, the (N-1) weight is 2, and the first weight is 1.