KR20110014758A - Fingerprint scan system and its liveness fingerprint detecting method - Google Patents

Abstract

PURPOSE: A fingerprint scan system and a method for detecting counterfeit fingerprint are provided to reduce the recognition false by using signal strength distribution and Fourier spectrum. CONSTITUTION: A fingerprint recognition module(20) scans and recognizes fingerprint, and generates the scanned fingerprint image. A fingerprint determining module(40) analyzes the fingerprint image in the spatial domain and frequency domain, so as to distinguish the counterfeit fingerprint. A spatial coordinate calculating module(50) produces the signal strength histogram index from the fingerprint image. A frequency index calculating module(60) extracts the Fourier spectrum from the fingerprint image, and produce the spectrum index.

Description

Fingerprint scan system and its liveness fingerprint detecting method

The present invention relates to a fingerprint recognition system and a counterfeit fingerprint identification method, and more particularly, a fingerprint recognition system and a counterfeit thereof, which can greatly improve the counterfeit fingerprint discrimination power compared to the related art even without a high-performance image sensor or an additional sensor. It relates to a fingerprint identification method.

1 is a diagram showing an example of a real fingerprint and a fake fingerprint, Figure 1 (a) is an image showing the actual fingerprint, Figure 1 (b) shows a fake fingerprint made from the real fingerprint of Figure 1 (a) Image.

Counterfeit fingerprints are made from materials such as silicone and gelatin, based on the actual fingerprints. As shown in FIG. 1 (b), such a fake fingerprint has almost the same characteristics of a fingerprint pattern as a standard for recognition, and thus it is impossible to identify the fake fingerprint with the fingerprint recognition system. Do.

Therefore, conventionally, a separate algorithm has been developed that applies a new sensor or a method for identifying a fake fingerprint as a method for identifying such a fake fingerprint.

In other words, as described above, the fake fingerprint can be made of various materials such as silicon, gelatin, film, etc., based on the actual fingerprint, and according to the material properties and the manufacturing process of the material, the signal distribution characteristics and the image quality are different. Because there is a difference in, it detects it and identifies fake fingerprints.

Among the methods for identifying a counterfeit fingerprint according to the prior art, a method for applying a new sensor includes applying a sensor such as a high resolution image sensor, an image sensor using light sources of various wavelengths, a light transmittance measurement sensor of a finger, and a capacitance measurement sensor. By applying a high-performance sensor can increase the fingerprint identification rate, but there is a problem in that high cost is consumed.

In addition, a method of identifying a fake fingerprint by using an output image signal from an existing image sensor includes a method of analyzing a signal distribution in a frequency domain after Fourier transform or a method of analyzing image quality of an input image by wavelet transform This has been proposed.

In the case of applying a Fourier transform, a two-dimensional Fourier spectrum is divided into two regions (Inner Ring and Outer Ring), and a fake fingerprint is identified by comparing both the energy of each region and the energy values of the entire region. In the experiment using one method, the probability of recognizing a real fingerprint as a fake fingerprint was 23%, the probability of recognizing a fake fingerprint as a real fingerprint was 12%, and the overall false recognition rate was 16%. There is a problem that the false recognition rate is quite high.

In addition, in the method using the wavelet transform, the extracted fingerprint image and the noise component of the reference image are extracted, and the extracted values are compared to identify the forged fingerprints. Since a high resolution image sensor of 1000 dpi or more is required, rather than an image sensor of 500 to 800 dpi level installed in the system, a considerable cost is additionally consumed, which is not suitable for commercialization.

The present invention is to solve the problems according to the prior art described above, by utilizing both the signal intensity distribution and the Fourier spectrum of the recognized fingerprint image to improve the reliability of the system by significantly lowering the false recognition rate of the fingerprint even without adding a new sensor The purpose of the present invention is to propose a fingerprint recognition system and a fake fingerprint identification method.

Fingerprint recognition system according to an embodiment of the present invention for this purpose, the fingerprint recognition module for recognizing and scanning the fingerprint and generating it as image data; And a fingerprint identification module for determining whether the fingerprint is a false fingerprint by analyzing the scanned fingerprint image in the spatial domain and the frequency domain, thereby greatly improving the discriminating power of the fake fingerprint without additionally adding a high-performance sensor.

In addition, the forged fingerprint identification method of the fingerprint recognition system according to an embodiment of the present invention, calculating the histogram indicator and the spectral indicator for the fingerprint image generated from the recognized fingerprint; Determining whether a calculated value of each index falls within a range of a histogram index and a spectrum index of a preset reference image; And determining the generated fingerprint image as a fake fingerprint when the calculated value does not belong to the set index range.

Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings.

2 is a block diagram showing the configuration of a fingerprint recognition system according to an embodiment of the present invention.

The fingerprint recognition system according to an embodiment of the present invention, as shown in FIG. 2, includes an image sensor 10, a fingerprint recognition module 20, a fingerprint DB 30, and an output unit 70.

When the user's finger or the like is touched, the image sensor 10 may scan the fingerprint of the contact portion and generate the image (image), and the fingerprint recognition module 20 registers or registers the actual fingerprint of each user and then the image sensor. By comparing the fingerprint image generated through the registered fingerprint image stored in the fingerprint DB 30 to determine whether or not the normal user. In addition, the output unit 70 outputs a message indicating whether the user is a normal user in the fingerprint recognition module 20 or a fake fingerprint identification in the fingerprint identification module 40 so that the user can recognize the result of the fingerprint recognition. do.

A fingerprint identification module 40, a spatial index calculation module 50, and a frequency index calculation module 60 for determining whether the input fingerprint image is a fake fingerprint are included.

That is, comparing fingerprint patterns is difficult to identify fake fingerprints. Therefore, it is desirable to quantify the quality difference between the actual fingerprint image and the fake fingerprint image and use it as an index. Accordingly, in the present invention, forgery fingerprints are discriminated by using the distribution difference in the spatial domain and the distribution difference in the frequency domain for the fingerprint image as two identification indicators. For this, the fingerprint identification module 40 and the spatial index calculation are performed. Module 50, and frequency indicator calculation module 60.

In the fingerprint recognition system according to an embodiment of the present invention, the spatial index calculation module 50 calculates the difference between the signal intensity histogram distribution between the input image and the reference image as an index, and the frequency index calculation module 60 inputs the index. The difference in Fourier spectral energy between the image and the reference image is calculated as an index. In this case, the reference image represents an average of actual fingerprint images stored in the fingerprint DB 30. In addition, the fingerprint identification module 40 receives each index value and determines whether it is within the actual fingerprint index range, and determines whether or not to forge accordingly.

In more detail, the spatial domain indicator expresses an input image as a histogram of 16 sections based on the intensity of a signal of each pixel, and then adds the histogram value (frequency) of 16 sections to each fingerprint. The difference between the reference image and the input image can be calculated by the following equation.

In the above formula, E is a histogram distribution difference between the input image X and the reference image R, p is a fingerprint set index (1 to 16), n is a fingerprint image index (1 to 50), and j is an input image. Interval indexes 1 to 16 of the histogram of the reference image.

In the above formula, the numerator corresponds to Mean Square Error (MSE), and the denominator corresponds to a kind of normalization factor added to increase discrimination. The reason for applying the normalization factor is that the values of the input image and the reference image may be relatively small in a predetermined section than the values in the other section. Although the difference in the values is relatively small, the rate of change with respect to the reference image is significant. This may be corrected by dividing by the value of the reference image.

Therefore, after obtaining the spatial region index of each fingerprint image through the above equation, it is determined whether the input image is a fake fingerprint by whether the spatial region index belongs to the actual fingerprint index range set for each fingerprint.

The width of the indicator range for judging the actual fingerprint greatly affects the false recognition rate. When the wide range of the actual fingerprint indicator range is set, the actual fingerprint false recognition rate that recognizes the actual fingerprint as a counterfeit fingerprint decreases, but the false fingerprint is recognized as the actual fingerprint. Counterfeit fingerprint misrecognition rates can increase. In addition, setting a narrow range of the actual fingerprint index may increase the actual fingerprint misrecognition rate but decrease the false fingerprint misrecognition rate. In an embodiment of the present invention, the width of the actual fingerprint index range may be set to about three times the standard deviation so that the actual fingerprint false recognition rate is within 2%.

Defining the width of the actual fingerprint index range to about three times the standard deviation corresponds to the range including 98% or more of the input signal on the normal distribution, and the basic data for calculating the standard deviation is the index value calculated from the actual fingerprint images. It is desirable to be utilized.

In order to adjust the false recognition rate, first adjust the range of the actual fingerprint index range to set the actual fingerprint false recognition level lower than the desired level, and then add an indicator to improve the false fingerprint false recognition rate.

The frequency domain analysis method is similar to the spatial domain analysis method described above, but since the Fourier spectrum in the frequency domain is a two-dimensional matrix signal, it is necessary to simplify it to a one-dimensional vector signal. This is to simplify the computational throughput in the index calculation process, and it may be difficult to process the 2D matrix signal due to system resource or processing time constraints. For example, if you want to experiment with forgery fingerprint identification by preparing 50 input images for each fingerprint for 16 fingerprint sets that use a real fingerprint and two kinds of fake fingerprints as one fingerprint set, convert it into a one-dimensional vector. When calculating the frequency index, the amount of calculation is about 576, but when calculating the frequency index from the two-dimensional vector, about 82,944 operations must be performed. This is not preferable because it consumes a lot of system resources and slows down the processing speed, which is about 144 times larger than that of the one-dimensional vector.

In the frequency domain, a method of converting a Fourier spectrum signal into a one-dimensional vector signal is obtained by sampling only two horizontal and vertical vector signals passing through the center from the Fourier spectrum signal, and then taking the average of the two to input an input to calculate an index. The signal was constructed.

On the other hand, the Fourier spectrum signal, which is simplified as a one-dimensional vector signal, is divided into 16 sections, and the index in the frequency domain is calculated by calculating the difference in Fourier spectrum energy between the reference image and the input image for the spectrum values of the 16 sections for each fingerprint. do. The equation for calculating the frequency index may have the same form as the equation for calculating the spatial domain index as follows, and whether or not the counterfeit fingerprint is included in the actual fingerprint index range as in the above-described spatial domain index. Can be determined.

In the above equation, E is a frequency domain index as the difference between the Fourier spectrum signal X of the input image and the Fourier spectrum signal R of the reference image, p is a fingerprint set index (1 to 16), and n is a fingerprint image index (1 to 50). ), j is the interval index 1 to 16 of the Fourier spectrum.

The spatial domain index and the frequency domain index calculated as described above may be used as an index for identifying a fake fingerprint individually, but both indexes may be combined to identify whether a fake fingerprint is present.

Since the areas identified by the counterfeit fingerprints are different from each other, the false recognition rate can be lowered when they are collectively identified. In this case, if the real fingerprint is falsely recognized as a counterfeit fingerprint, it is the union between the two indices, so the overall real fingerprint misrecognition rate is increased than the respective real fingerprint misrecognition rate. Therefore, the total number of false fingerprint false recognition rates is reduced than that of each false fingerprint false recognition rate.

Thus, the fingerprint recognition system according to an embodiment of the present invention can improve the false recognition rate of the fake fingerprint by adding a new indicator in addition to the two indicators applied in the present specification. Of course, when two indicators are applied, it can be improved only by applying indicators that can be correctly identified for the images that are incorrectly recognized. The overall false positive rate can be further improved.

On the other hand, in the fingerprint recognition system according to an embodiment of the present invention, the fingerprint identification module 40, the spatial index calculation module 50, and the frequency index calculation module 60 for the fake fingerprint identification is one of these modules It can be made as a separate module, in this case it is possible to identify the fake fingerprint in accordance with an embodiment of the present invention by adding to the existing fingerprint recognition system.

3 is a flowchart illustrating a method for identifying a fake fingerprint in a fingerprint recognition system according to an embodiment of the present invention.

First, in step S10, an actual fingerprint is registered for each user.

In operation S20, when the fingerprint is recognized and a fingerprint image is generated, the spatial domain index and the frequency domain index are calculated in the same manner as described above with respect to the fingerprint images generated in operations S30 and S40.

In step S50, the fingerprint DB is searched for the generated fingerprint image to determine whether the same fingerprint image as the fingerprint pattern of the generated fingerprint image exists in the fingerprint DB, and if the same fingerprint image exists, in step S60, the generated fingerprint The spatial domain index and the frequency domain index of the image and the reference image are compared to determine whether they correspond to the actual fingerprint recognition range.

As a result of the determination, when it is determined that the fingerprint is included in the actual fingerprint recognition range and is the actual fingerprint, it is recognized as a normal user in step S80, and it is approved as a normal user through the output unit.

In addition, if the generated fingerprint image is not included in the actual fingerprint recognition range as a result of the determination, in step S70, it is determined that the fingerprint is a fake fingerprint and outputs a message indicating that the fingerprint is fake.

In the above-described steps, steps S30 and S40 of calculating the spatial domain index and the frequency domain index may be performed when it is determined in step S50 that the fingerprint is a registered fingerprint and then whether a fake fingerprint is present.

In operation S50, the determining whether the generated fingerprint image coincides with the registered fingerprint image is to determine whether the fingerprint is fake after fingerprint recognition, and all users are compared with a method of performing a fingerprint recognition process after determining the fake fingerprint. By narrowing the range to a specific user, you can compare image features to speed up processing and reduce system resources.

Forgery fingerprint identification method according to an embodiment of the present invention because only the fingerprint image as an input signal, it may be implemented in software and added to the existing fingerprint recognition system.

Experiments for verifying the fake fingerprint identification capability of the fingerprint recognition system and the fake fingerprint identification method according to an embodiment of the present invention constructed and operated as described above were performed as follows.

The fingerprint DB used in the false recognition rate experiment consisted of 16 fingerprint sets, each consisting of a real fingerprint, a fake fingerprint made of silicon, and a fake fingerprint made of gelatin for the same finger. In addition, 50 image samples were provided for each fingerprint, and the size of the fingerprint image was 288X 288 pixels, which was generated by an image sensor used in a commercially available fingerprint recognition system. Each fake fingerprint image is recognized as a real fingerprint in the existing fingerprint recognition system, and the total number of fingerprint images is 16X3X50 = 2400.

4 is an exemplary diagram illustrating a distribution of spatial domain indices in a fingerprint recognition system according to an exemplary embodiment of the present invention.

As described above, the histogram according to the signal strength of each fingerprint image may be applied to a formula to calculate the spatial region index of the fingerprint image.

In Figure 4, the horizontal axis is the number of 50 fingerprint images, the circle (H1) is the indicator value of the actual fingerprint, the asterisk (H2) is the indicator value of the silicon fake fingerprint, the square (H3) is the indicator value of the gelatin fake fingerprint.

The first and second solid lines L1 and L2 around the circle represent the upper and lower limits of the actual fingerprint index range and represent the actual fingerprint index range. According to FIG. 4, since the index values of the fake fingerprints used in the experiment are all outside the actual fingerprint index range, all of them were correctly determined as fake fingerprints. However, two of the actual fingerprint images were misidentified as fake fingerprints outside the range of the actual fingerprint index.

4 is a diagram illustrating a case where the index value of the actual fingerprint and the index value of the fake fingerprint are relatively well separated, and thus may not appear similarly to FIG. 4 in all cases.

In order to see how well the counterfeit fingerprint can be identified by applying the above-described spatial domain indicator, the average value of the spatial domain indicator is calculated for 50 images of real fingerprints, silicon fake fingerprints and gelatin fake fingerprints for each of 16 fingerprint sets. It was shown as a graph.

FIG. 5 is an exemplary diagram illustrating an average value distribution of the spatial domain indicators illustrated in FIG. 4 in the fingerprint recognition system according to an exemplary embodiment of the present invention.

In Fig. 5, the horizontal axis is a fingerprint set number, a circle M1 is an indicator value of an actual fingerprint, an asterisk M2 is an indicator value of a silicon fake fingerprint, and a square M3 is an indicator value of a gelatin fake fingerprint. As shown in FIG. 5, since both the index values M2 and M3 of the fake fingerprints are larger than the index values M1 of the actual fingerprints, it is determined that the spatial domain index proposed in the present invention is a high discriminating index. False recognition rates can be expected.

As shown in the graph of FIG. 5, it can be seen that a histogram distribution difference clearly exists between the actual fingerprint image and the fake fingerprint image. Even when the actual fingerprint image and the fake fingerprint image are visually compared, the fake fingerprint image is generally raised and valleys. The transition between them is not smooth and sharp, and the seams of the ridges are often broken. Therefore, it can be seen that these differences, such as the frequency of the mid-intensity signal or the loss of some signals, are reflected in the spatial domain indicator, and the differences that can be distinguished by the naked eye must be clearly identified first of all. The need to apply spatial domain indicators is supported.

In the false recognition rate experiment conducted using only the spatial domain indicator, the total number of fingerprints used in the experiment was 48 (16 actual fingerprints, 16 silicon fake fingerprints, 16 gelatin fake fingerprints), and 50 fingerprints for each fingerprint were 2,400. False recognition rates were calculated for dog input images.

Real fingerprint Silicone fingerprint Gelatin fingerprints Full fingerprint False recognition count () 16 67 61 144 False recognition rate (%) 2 8 8 6 Recognition rate (%) 98 92 92 94

As shown in Table 1, the actual fingerprint false recognition rate was 2% because the width of the actual fingerprint index area was set to 3 times the standard deviation, and the false fingerprint false recognition rate was 8%. The recognition rate was 94%, indicating a low false recognition rate.

The average value of the frequency domain indicator for the 50 input images of the real fingerprint, the silicon fake fingerprint, and the gelatin fake fingerprint for each of the 16 fingerprint sets, as in the case of the spatial domain indicator, is used for the identification of the false fingerprint for the frequency domain indicator. Was plotted.

6 is an exemplary diagram illustrating an average value distribution of a frequency domain indicator in a fingerprint recognition system according to an embodiment of the present invention.

In the graph of FIG. 6, a circle D1 represents an index value of an actual fingerprint, an asterisk D2 represents an index value of a silicon fake fingerprint, and a square D3 represents an index value of a gelatin fake fingerprint. According to FIG. 6, the index value of some of the fake fingerprints is close to the index value of the actual fingerprint, which is highly likely to be misidentified. However, the indicator value of the fake fingerprint is larger than the index value of the actual fingerprint. can do.

The results of the false recognition rate experiments performed using only the frequency domain indicators are shown in Table 2 below, and the fingerprints used in the experiments are the same as the spatial domain indicators.

Real fingerprint Silicone fingerprint Gelatin fingerprints Full fingerprint False recognition count () 19 294 221 534 False recognition rate (%) 2 37 28 22 Recognition rate (%) 98 63 72 78

As a result, as shown in Table 2, the actual fingerprint false recognition rate was maintained at 2% because the width of the actual fingerprint index area was set to 3 times the standard deviation, but the false recognition rate of the fake fingerprint was incorrect when only the spatial area index was applied. Compared to the recognition rate, 8% was 33%. The overall fingerprint recognition rate is 78%, which is relatively lower than 94%, the total fingerprint recognition rate when only the spatial domain index is applied. However, since the same fingerprint set shows different false recognition rates, it can be determined that the fake fingerprints are identified from different viewpoints. Therefore, the frequency domain indicator is preferably applied in view of complementing the spatial domain indicator for various counterfeit fingerprints.

The results of conducting the fake fingerprint identification experiment using both spatial domain and frequency domain indicators are shown in Table 3 below. The fingerprints used in the experiments are the same as those used in the spatial domain and frequency domain indicator experiments.

Real fingerprint Silicone fingerprint Gelatin fingerprints Full fingerprint False recognition count () 29 53 35 117 False recognition rate (%) 4 7 4 5 Recognition rate (%) 96 93 96 95

Experimental results showed that the actual fingerprint misrecognition rate increased by about 2% due to the union effect, while the false fingerprint misrecognition rate decreased by about 2.5% due to the intersection effect, and the overall recognition rate was about 1%. It appears to have increased. The reduction of the false fingerprint misrecognition rate is due to the complementary effect of the frequency domain index, and it can be seen that the fake fingerprint, which is difficult to identify with the spatial domain index, can be identified through the frequency domain index.

As described above, the fingerprint recognition system and the counterfeit fingerprint identification method according to the present invention have been described with reference to the illustrated drawings, but are not limited to the embodiments and drawings disclosed in the present specification. It is apparent that the technical idea of the present invention, which calculates a frequency domain index and compares the index with the index of the reference image, respectively, to improve the false fingerprint identification rate, can be easily applied by those skilled in the art within a protected range.

1 is a diagram showing a real fingerprint and a fake fingerprint;

2 is a block diagram showing the configuration of a fingerprint recognition system according to an embodiment of the present invention;

3 is a flow chart showing a fake fingerprint identification method in the fingerprint recognition system according to an embodiment of the present invention,

4 is an exemplary diagram illustrating a distribution of spatial region indices in a fingerprint recognition system according to an embodiment of the present invention;

FIG. 5 is an exemplary diagram illustrating an average value distribution of a spatial domain index shown in FIG. 4 in a fingerprint recognition system according to an embodiment of the present invention; and

6 is an exemplary diagram illustrating an average value distribution of a frequency domain indicator in a fingerprint recognition system according to an embodiment of the present invention.

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

20: fingerprint recognition module 30: fingerprint DB

40: fingerprint identification module 40: space index calculation module

50: frequency index calculation module 60: fingerprint DB

Claims (8)

A fingerprint recognition module for recognizing and scanning a fingerprint and generating the fingerprint as image data; And Fingerprint identification module that analyzes the scanned fingerprint image in the space domain and frequency domain to determine whether it is a fake fingerprint Fingerprint recognition system comprising a. The method of claim 1, The fingerprint identification module comprises: a spatial coordinate calculation module for dividing the scanned fingerprint image into a plurality of sections based on the signal strength of each pixel and calculating a signal intensity histogram index in each section; And A frequency index calculation module for extracting a Fourier spectrum from the scanned fingerprint image and dividing it into a plurality of sections, and calculating a spectral index by calculating the spectral energy difference between the reference image and the scanned fingerprint image for each section. Fingerprint recognition system comprising a. The method of claim 2, The plurality of sections are fingerprint recognition system consisting of 16 sections. The method of claim 2, The fingerprint identification module sets a range of the histogram index and the spectral index for the actual fingerprint, and if the histogram index and the spectral index for the scanned fingerprint image do not belong to the set range, respectively, the scanned fingerprint image is a fake fingerprint. Fingerprint identification system to judge. The method of claim 4, wherein The fingerprint identification module is configured to set the width of the range of the histogram indicator and the spectral indicator to three times the standard deviation for each indicator. Calculating a histogram indicator and a spectral indicator on the fingerprint image generated from the recognized fingerprint; Determining whether a calculated value of each index falls within a range of a histogram index and a spectrum index of a preset reference image; And Determining the generated fingerprint image as a fake fingerprint if the calculated value does not belong to the set index range; Counterfeit fingerprint identification method of the fingerprint recognition system comprising a. The method of claim 6, Computing the histogram indicator and the spectral indicator, Deriving a histogram index by dividing the generated fingerprint image into a plurality of sections according to signal strength and calculating a signal intensity histogram distribution difference with the reference image; Converting the generated fingerprint image into a Fourier spectrum and calculating a spectral energy difference from the reference image to derive a spectral index Counterfeit fingerprint identification method of the fingerprint recognition system comprising a. The method of claim 6, Computing the spectral indicators, Sampling two vector signals from a Fourier spectrum signal of the generated fingerprint image, calculating an average, and converting the two vector signals into a one-dimensional vector signal; Counterfeit fingerprint identification method of the fingerprint recognition system further comprising.

Images