KR20070076187A - Fingerprint Identification Method - Google Patents

Abstract

A method for recognizing a fingerprint is provided to precisely recognize ridge information of the fingerprint and remove disconnection of the ridge in which the fingerprint is severely changed by performing interpolation. A core area of the fingerprint is extracted by dividing a fingerprint area into each block, detecting the ridge direction/frequency of each block, and using the detected ridge direction/frequency of each block(120). A difference of ridge directions between blocks is calculated for the blocks included in the core and neighboring area of the fingerprint(130). The interpolation of the ridge direction is performed if the calculated difference between the ridge directions is over a threshold value(150). The interpolation is performed by dividing the difference of the ridge directions into the number of sub blocks and adding or subtracting the difference from each ridge direction. Gabor filtering is performed by using the interpolated ridge direction(160).

Description

Fingerprint recognition method

1 is a flow chart showing a fingerprint recognition method according to the present invention.

2 shows an example of fingerprint filtering in a gabor filter.

3 is a block diagram of a fingerprint image.

4 shows an example of a Sobel edge calculation.

5 is a diagram illustrating a conventional gigabit filtering result.

6 is a flowchart illustrating a fingerprint recognition method according to the present invention.

7 illustrates an example of interpolation in the inter-block fingerprint ridge direction according to the present invention.

8 is a diagram showing a result of gabor filtering after fingerprint ridge interpolation according to the present invention.

The present invention relates to a fingerprint recognition method in which ridge interpolation can be performed, particularly in areas where fingerprint changes are severe.

As the importance of privacy and security is emphasized, much research and attention has been focused on biometric techniques such as fingerprint, retina, iris, and face recognition. Among them, the fingerprint authentication method using the uniqueness of each individual's fingerprints does not overlap, which is superior to other biometric methods in terms of reliability, stability, and processing speed. It is recognized as a means.

There are a number of feature regions in the fingerprint, in addition to the normal region of directional fingerprint ridges. In this feature, there is a lot of lighting in the fingerprint, especially the ridge proceeds and the ridge proceeds and the lighting is separated. The ridge proceeds and the boiling point is called an ending point, and the ridge splits. Is referred to as a bifurcation point, which is collectively referred to as a minutiae of the fingerprint. Generally, about 100 to 150 of these feature points are distributed on one finger, and each person is known to have a different type, position, and direction. Therefore, the information on the feature point of the fingerprint can be used as an identification means for each individual, and if the positional relationship and the direction of the feature point match 50 to 60% or more, it is usually determined to be the same fingerprint.

Fingerprint authentication in fingerprint recognition system basically extracts minutiae such as Bifurcation Point and Ending Point from each person's fingerprint, and their absolute position information (X, Y Coordinate) or relative position information And, by using the information (Angle), etc. that these feature points make a template (Template) and is made by matching between the fingerprint that is already stored and registered and the fingerprint entered for authentication. Such a fingerprint recognition system has recently been applied to various kinds of devices, and especially has been introduced to portable terminals.

However, when users enter their fingerprints for authentication, it is not possible to enter them at the same intensity, same rotation angle, and same location as when enrolling the fingerprints, thus denying their identity resulting from Misunderstanding problems such as acceptance of others are frequently occurring.

Accordingly, there is a need for acquiring more accurate information in the situation of fingerprint input or fingerprint image, that is, ridge direction information and continuity of the ridge, and improving the fingerprint recognition rate.

It is a first object of the present invention to provide a fingerprint recognition method for recognizing a fingerprint.

It is a second object of the present invention to provide a fingerprint recognition method for recognizing ridge information of a fingerprint more accurately.

A third object of the present invention is to provide a fingerprint recognition method capable of eliminating the boiling phenomenon of the ridges in a region where the change in the orientation information of the fingerprint is severe.

Fingerprint recognition method according to the present invention for achieving the above object,

Extracting a central region of the fingerprint;

Calculating inter-block fingerprint ridge direction differences for blocks included in the central area and the adjacent area of the fingerprint;

And performing inter-block ridge interpolation if the calculated ridge direction difference exceeds a threshold.

Preferably, the extracting of the fingerprint center region comprises: dividing the fingerprint region into blocks; Detecting ridge direction and frequency for each block; And extracting a central region of the fingerprint using the ridge direction and the frequency of each block.

Preferably, the central region of the fingerprint is characterized by using a Poincare index (Poincare index).

Preferably, the inter-block fingerprint ridge interpolation may include applying a representative direction vector of each sub-block obtained by dividing the ridge direction difference value between the blocks into sub-blocks and then subtracting and subtracting the ridge direction. .

Preferably, the method further comprises performing gabor filtering using the interpolated ridge direction.

Preferably, the fingerprint ridge direction difference may be performed in at least two block units.

Referring to the accompanying drawings, a fingerprint recognition method according to an embodiment of the present invention configured as described above is as follows.

The general fingerprint recognition process consists of a preprocessing process for extracting feature points of a fingerprint, a direction pattern and feature point extraction process used for matching and discrimination, and a process of identification and recognition using them. In addition, since the fingerprint image is not always located at the same position and is acquired and moved or rotated in a constant direction, or the size is changed, simple comparison is impossible. In the matching process, an alignment step of aligning coordinates of two fingerprints is essential.

There are a number of feature regions in the fingerprint, in addition to the normal region of directional fingerprint ridges. In this feature, there is a lot of lighting in the fingerprint, especially the ridge proceeds and the ridge proceeds and the lighting is separated. The ridge proceeds and the boiling point is called an ending point, and the ridge splits. Is referred to as a bifurcation point, which is collectively referred to as a minutiae of the fingerprint. Generally, about 100 to 150 of these feature points are distributed on one finger, and each person is known to have a different type, position, and direction. Therefore, the information on the feature point of the fingerprint can be used as an identification means for each individual, and if the positional relationship and the direction of the feature point match 50 to 60% or more, it is usually determined to be the same fingerprint.

A fingerprint recognition method will be described with reference to FIG. 1.

When a fingerprint is input (10), a preprocessing process is performed (20). The preprocessing process normalizes the fingerprint with respect to the input fingerprint image (21), and after detecting a fingerprint region (22), the fingerprint ridge direction / frequency It is detected (23). In this case, if the direction and frequency of the fingerprint ridge are detected, gabor filtering is performed (24), and the fingerprint is binarized (25).

Here, in the fingerprint binarization process, a process for reducing the amount of data to be handled, and the binarization of the fingerprint image refers to a process of distinguishing the ridge portion, which is the protrusion of the fingerprint, and the curved portion, which is the fine portion, in black and white. Since the input image may include a background region other than the fingerprint, the input image may include a process of distinguishing the fingerprint region from the background region.

After the preprocessing process 20, the fingerprint thinning 30 and the fingerprint feature vector extraction 40 perform matching with the fingerprint feature information stored in the database 60 to check whether the fingerprint is the same. do.

Specifically, the fingerprint thinning process 30 means generating a line image having a width of a ridge of one pixel from a binary image, and the thinned fingerprint image is advantageous in obtaining a feature point. In the correction process, the pseudo feature points appear in the set of feature points due to the conditions or processing limitations when the fingerprint image is acquired. The pseudo feature points are removed from the thinned image.

In the feature vector extraction process 40, a feature point existing in the fingerprint is extracted. In the matching process 50, the fingerprint is identified and recognized using the feature points extracted in the above process. Step.

Fingerprint recognition is performed in this manner. In general, there are scars, wounds, sweat glands, and the like, and dry fingerprints, wet fingerprints, and the like appear depending on the input environment. Therefore, the preprocessing process is an essential process in the fingerprint recognition method, and one of the methods used here is mainly used for the gabor filtering process.

The gabor filter is defined as shown in Equation 1, and is represented as shown in FIG.

이고,

이며,

는 게이버 필터의 방향이고, f는 주파수이다.

는 x 및 y축으로 가우시안(Gaussian)면의 공간 상수이다.here,

ego,

Is,

Is the direction of the gabor filter and f is the frequency.

Is the spatial constant of the Gaussian plane on the x and y axes.

Generally, the ridge direction of the fingerprint is calculated by dividing the fingerprint area into M * M (or M * N) block units as shown in FIG. 3 due to the characteristics of the fingerprint area. In order to obtain the ridge direction, first, the gradient images Gx and Gy are obtained using a sobel edge mask as shown in FIG. 4. For this image, the ridge directions Vx and Vy are calculated using Equation 2 below.

Here, Gx and Gy are Sobel edge result values in each pixel. W represents a block size, and θ represents a representative direction value of each block. In addition, in order to minimize the influence of fingerprint miscellaneous, the total 360 degrees are quantized in 45 degree units.

The result of filtering the fingerprint image by applying the representative direction value to the direction component of the gabor filter of Equation 1 is shown in FIG. 5. Good results can be obtained when the inter-block variation of the fingerprint ridge direction information is small, but an error can be generated when the inter-block fingerprint ridge direction information such as the core point of the fingerprint is large.

In other words, the direction information and the frequency component of the fingerprint ridge are the main variables used in the gabor filter in improving the fingerprint image quality. These two variables apply the representative value calculated in each area divided into blocks to all the pixels included in the block. However, when the inter-block ridge information variation, such as the center region, is large, that is, in the fingerprint having a large curvature, the result of the gabor filter appears discontinuously.

An example of such an error may occur that the filtered ridge direction in the region P of FIG. 5 is interrupted. Such a problem may cause an increase in fingerprint recognition time, and may eventually cause a decrease in fingerprint recognition rate. As described above, in the region having a large curvature of the fingerprint, the direction information of the fingerprint ridge is severely changed. If the gabor filtering is performed on the region, the ridge break occurs. That is, discontinuities in the fingerprint ridges may appear.

6 is a flowchart illustrating a fingerprint recognition method according to the present invention.

Referring to FIG. 6, first, the fingerprint ridge direction and the frequency of each block are detected in the preprocessing process (20 of FIG. 1), and then the core region of the fingerprint using the fingerprint ridge information and the frequency information. It will be extracted (120). The inter-block fingerprint ridge direction difference is calculated with respect to the blocks included in the extracted central area and the surrounding area (130). If the ridge direction difference is higher than the threshold a, inter-block ridge direction interpolation is performed (150). The ridge direction difference is smaller than the threshold value, or the inter-block ridge direction is performed for the gabor filtering 160 for the region.

Specifically, for each step,

Extracting the fingerprint center region 120 uses a Poincare index (P) as shown in Equation 3 as a method for extracting the fingerprint region.

Here, d is ridge direction information. If P (i, j) is 180 degrees, it corresponds to the center region.

After the operation 120, the block ridge direction difference between the blocks is calculated with respect to the blocks included in the extracted central area and the surrounding area (130). Here, the fingerprint ridge direction difference between blocks is calculated with respect to 6 or 8 around the center area block. Here, the central region and the peripheral region are obtained as a set of blocks, and the region means one pixel unit.

The direction of the ridges for each block is to calculate the difference between the ridge direction of the two blocks on the basis of a predetermined direction (130). In this case, the ridge direction of each block is determined by Equation 2 above, and the slope corresponding to the ridge direction is a value that adds up and averages the number of ridges in the block.

Then, the ridge direction difference value between the two blocks is compared with the threshold value t. If the ridge difference value between the blocks is greater than the threshold value t, the fingerprint ridge interpolation between the two blocks is performed (150).

)가 임계치(t) 보다 크면 두 블록간에 지문 융선 보간이 수행된다. As shown in FIG. 7, the absolute value of the difference between the slope a of the first block W1 and the slope b of the second block W2 (

Is greater than the threshold t, fingerprint ridge interpolation is performed between the two blocks.

The ridge direction (tilt) for each block is an average slope value of the ridges in the block, and the ridge directions of two adjacent blocks calculate the slope in the same clockwise direction. Here, the adjacent directions may be between horizontal blocks, between vertical blocks, or between diagonal blocks.

When one block W1 or W2 is 8 * 8, the inter-block fingerprint ridge interpolation pixel is 4 * 8 for each block. The direction difference may be calculated for each of the 1/2 blocks W1-5 to W1-8 and W2-5 to W2-8 adjacent to the first block W1 and the second block W2.

Since two blocks are performed, 8 * 8 pixels corresponding to one block W3 are obtained. As another example, one block may be n * n (eg, 16 * 16) or n * m (eg, 8 * 16).

As described above, the ridge direction information D newly applied to the ridge direction interpolation pixels of each block is expressed by Equation 4 below.

Here, D is ridge direction information, i (e.g., i = 1, ..., 8) represents each pixel, and W is 8, since W3 = 8 * 8 as a block size (or number of pixels). In this way, the slope is gradually connected to the ridge slope of the adjacent block by gradually adding or decreasing the slope for each subpixel.

In other words, by dividing the difference between a and b by 8 (= block size), the equal angle is continuously added or subtracted based on the a or b value to obtain a representative direction vector of each subblock, and the obtained value is included in the block. The same applies to all the pixels. The angles a and b for each block are viewed to be + or-depending on the situation so that the angle smoothly changes between a and b.

As described above, by performing interpolation between the ridge direction differences between the blocks, and performing gabor filtering, a gabor-filtered image as shown in FIG. 8 appears. In particular, when the fingerprint center region P 'is compared with the fingerprint center region P of FIG. In FIG. 8, the fingerprint ridge break is eliminated.

In the present invention, if the central region is present in one block, if the block is 8 * 8, ridge interpolation in the block is removed by performing ridge interpolation of the size of 4 * 8 or 8 * 4. In addition, when a plurality of central regions exist, fingerprint ridge interpolation is performed for each candidate block by two blocks. In addition, the fingerprint ridge interpolation between blocks may be prevented in the delta region in which the fingerprint is severely changed as described above.

So far, the present invention has been described with reference to the preferred embodiments, and those skilled in the art to which the present invention pertains to the detailed description of the present invention and other forms of embodiments within the essential technical scope of the present invention. Could be implemented. Here, the essential technical scope of the present invention is shown in the claims, and all differences within the equivalent range will be construed as being included in the present invention.

According to the fingerprint authentication method according to the present invention, after extracting the center region of the fingerprint, after calculating the ridge direction difference between the blocks for the blocks included in the fingerprint center area and the peripheral area, if the ridge direction difference is higher than the threshold between the blocks By performing ridge interpolation, there is an effect of eliminating the problem of breaking the fingerprint ridge.

In addition, the fingerprint ridge is prevented from breaking, thereby improving the fingerprint recognition rate.

In addition, the fingerprint ridge break can be eliminated in advance, thereby reducing the processing time for fingerprint recognition.

Claims (6)

Extracting a central region of the fingerprint; Calculating inter-block fingerprint ridge direction differences for blocks included in the central area and the adjacent area of the fingerprint; And performing inter-block ridge direction interpolation if the calculated ridge direction difference exceeds a threshold. The method of claim 1, The fingerprint center region extraction step, Dividing the fingerprint area into blocks; Detecting ridge direction and frequency for each block; And extracting a center area of the fingerprint using the ridge direction and the frequency of each block. The method of claim 2, And a center region of the fingerprint using a Poincare index. The method of claim 1, The inter-block fingerprint ridge interpolation step may include; And dividing the ridge direction difference value between the blocks into subblocks, and then applying the representative direction vector of each subblock, which is obtained by subtracting and subtracting each ridge direction, to all pixels. The method of claim 1, And performing gabor filtering using the interpolated ridge direction. The method of claim 1, The fingerprint ridge direction difference is a fingerprint recognition method, characterized in that performed in at least two blocks.

Images