KR100904902B1 - Method and apparatus for determining forged fingerprint for optical fingerprint acquisition apparatus - Google Patents

Abstract

The present invention relates to a method for discriminating a biometric fingerprint and a fake fingerprint. First, feature values are extracted from absorbed images and scattered images acquired from a plurality of biological fingerprints to obtain distribution characteristics (eg, mean and variance) in a feature space. Similarly, the distribution characteristics are obtained in the same way for a large number of fake fingerprints. When the distribution characteristics of these two types of fingerprints are obtained, a pattern clasifier for classifying the two patterns can be constructed using various pattern recognition techniques. When an unknown fingerprint is input, the designed pattern classifier may be used to determine whether the unknown fingerprint is a biometric fingerprint or a fake fingerprint.

Description

Method and apparatus for identifying fingerprint of optical fingerprint image acquisition device {Method and apparatus for determining forged fingerprint for optical fingerprint acquisition apparatus}

According to the present invention, a live fingerprint input by a living person's hand and an artificial fingerprint (forged fingerprint) imitating the fingerprint image, an absorption method and a scattering method are used. The present invention relates to a method and apparatus for distinguishing fingerprints that can be identified by using an optical fingerprint acquisition apparatus.

As fingerprint recognition devices are widely used in personal authentication fields such as access authentication and payment, security enhancement is required. In order to enhance the accuracy of fingerprint recognition, methods to cope with fake fingerprints are being developed one after another, but accordingly, fingerprint replication technology is also developing.

There are several prior patents regarding the method of identifying the duplicated fingerprints. 1) Japanese Unexamined Patent Publication No. Hei 11-45338 discloses a technique for detecting a biopotential generated in a living body and recognizing a counterfeit fingerprint. However, the electrode for potential detection in a state in which a counterfeit fingerprint is in contact with a fingerprint input window is disclosed. However, in case of contact with the living body, the imitation fingerprint cannot be recognized, and additional hardware and arithmetic processing part must be provided because the detection signal processing and frequency analysis process are separately required. 2) Dong gong (Pyeong) 10-290796 discloses a method of identifying the imitation fingerprints by measuring the form of giving a variety of stimulation to the living body in response to the stimulus. However, the method itself for analyzing the response of the human body by external stimulation is artificial, it is not only unpleasant to the user, but also difficult to quantify the form of response to the stimulation regularly. 3) Korean Patent Publication No. 9-259272 discloses a method for identifying a fake fingerprint by detecting the presence and number of sweat glands in a fingerprint image. However, according to the current imitation fingerprint production technology, not only the fingerprint form but also the glands can be copied as it is not realistic. 4) Korean Patent Publication No. 7-308308 discloses a technique for identifying a fake fingerprint by detecting changes in oxygen concentration and blood flow using a quantity of light transmitted by irradiating a specific light beam. However, when the imitation fingerprint is made of a material that transmits well with respect to the frequency of the light source emitted from the light source element and is worn on the finger, there is a limitation that the imitation fingerprint cannot be identified. 5) In addition, a technique for identifying a fake fingerprint by a method of recognizing a pulse of a fingertip by using a pressure sensor has been developed, but in this case, there is a limitation in that it cannot cope with other input information similar to a pulse.

In addition to the above limitations, in recent years, a technique for manufacturing a thin imitation fingerprint made of a material such as silicone or film has appeared. When the bio fingerprint and the imitation fingerprint are applied by any of the above methods when directly attached to or worn on a biological finger. There is almost no possibility of distinguishing.

Another subject is an optical fingerprint image acquisition device as a kind of fingerprint image acquisition device. The device operates in either absorption or scattering mode, depending on its structure. 1 is a principle diagram of an absorption type fingerprint image acquisition device, and includes a light source 112, a prism 110, a lens unit 114, an image sensor 116, and an image processing unit 125. The inclined surface of the prism 110 is a fingerprint contact surface 118 through which the fingerprint to be acquired is contacted. The inner surface of the fingerprint input window 118 is a total reflection surface that causes internal total reflection. do. With no fingerprint applied to the fingerprint input window 118, the light from the light source 112 is totally reflected on the inner surface of the fingerprint input window 118 of the prism 110 to the image sensor 116 through the lens 114. Incident. When the fingerprint is placed on the fingerprint input window 118, as shown in FIG. 2, the valley 130 of the fingerprint is located away from the fingerprint input window 118 so that incident light toward the valley 130 ( 127 reaches the image sensor 116 as reflected light 128 totally reflected from the inner surface of the fingerprint input window 118, and the ridge 132 of the fingerprint contacts the fingerprint input window 118. Most of the incident light 127 ′ toward 132 is absorbed by the melt 132 instead of being totally reflected inside the fingerprint input window 118 so that only some of the reflected light 129 reaches the image sensor 116. Therefore, the amount of light incident on the image sensor 116 is different from each other depending on the valleys and melting of the fingerprint. In the image sensor 116, the electrical signals of different levels according to the pattern of the fingerprint (ie, the combination of valleys and melting) are different. Will print The image processor 125 processes and outputs the output value of the image sensor 116 to recognize the pattern of the fingerprint image.

3 shows a principle diagram of a scattering fingerprint image acquisition device. The configuration includes a light source 312, a prism 310, a lens unit 314, and an image sensor 316 similar to the case of FIG. 1, but unlike the absorption method of FIG. 1, light is emitted from the light source 312. Since the light enters the fingerprint input window 318 of the prism 310 at a right angle or a much smaller angle than the critical angle, the incident light 327 is generated at the valley of the fingerprint that does not touch the fingerprint input window 318 instead of the internal reflection. Passing through the fingerprint input window 318 does not reach the image sensor, the incident light 327 'is scattered in the melting of the fingerprint and the scattered light 329 is detected by the image sensor 316 through the lens unit 314 ( See FIG. 4).

As described above, in the absorption type fingerprint image acquisition device, since absorption of light occurs at the melting part, the focused fingerprint image on the image sensor is dark, and the valley is bright. On the other hand, in the scattering fingerprint image acquisition device, light scattering occurs at the melting part, so that the fingerprint image formed on the image sensor is bright and the valley is dark, resulting in a contrast-inverted fingerprint image. FIG. 5 illustrates a biometric fingerprint image (a) obtained by an absorption method and a biometric fingerprint image (b) obtained by a scattering method.

In the case of the scattering method, the characteristics of the image reflected from the melting portion of the fingerprint (for example, light quantity or gray level) vary depending on the material and color of the melting portion. For example, in a dummy fingerprint made of a material that reflects better than the molten portion of the bio fingerprint, the amount of reflected light larger than the amount of light reflected from the bio fingerprint is measured. Similarly, in the case of the absorption method, the characteristic of the image absorbed and formed in the melted portion of the fingerprint also shows a difference depending on the material and the color. For example, in the molten part of the imitation fingerprint made of a material having good light absorption, the reflected light of a smaller amount of light is measured than in the case of the biofingerprint because it absorbs a greater amount of light than the melting part of the biofingerprint.

In contrast to the example of the bio fingerprint image of FIG. 5, a fake fingerprint image acquired by an absorption method and a scattering fingerprint image acquisition device is illustrated in FIGS. 6A and 6B, respectively. Comparing FIG. 5A with FIG. 6A, in the case of the image acquired by the scattering method, it can be seen that the dummy fingerprint image is darker than the biofingerprint image. 5 and 6 are images received through the CMOS image sensor (image size: 260x300 pixel).

The present invention focuses on the differences in the characteristics of the images acquired by the scattering method and the absorption type fingerprint image acquisition device, and first extracts the feature values from the absorbed images and scattered images acquired from the plurality of paper fingerprints. Find the distribution characteristics (eg mean, variance, etc.) on the feature space. Similarly, the distribution characteristics are obtained in the same way for a large number of fake fingerprints. When the distribution characteristics of the two types of fingerprints are obtained, a pattern classifier for classifying the two patterns can be constructed using various pattern recognition techniques. Using the pattern classifier designed as described above, it is determined whether the input fingerprint is a biometric fingerprint or a fake fingerprint.

In order to realize the above problem, according to one feature of the present invention

Extracting feature values from absorbed images and scattered images acquired from a plurality of biological fingerprints to obtain distribution characteristics (eg, mean, variance, etc.) in a feature space;

Extracting feature values from absorption images and scattering images acquired from a plurality of imitation fingerprints to obtain distribution characteristics (eg, average, variance, etc.) in a feature space;

Constructing a pattern classifier for classifying or discriminating two distribution patterns using various pattern recognition techniques for the distribution characteristics obtained for these two types of fingerprints;

Acquiring a first image by an absorption method and a second image by a scattering method from a fingerprint contacted with the fingerprint image acquisition device for fingerprint authentication;

Extracting a feature value by analyzing the first image and the second image, and obtaining the output of the pattern classifier using the feature value as an input of the pattern classifier;

There is provided a method of determining a biometric fingerprint and an imitation fingerprint in an optical fingerprint image acquisition device, comprising determining whether an object in contact with the fingerprint image acquisition device is a biometric fingerprint or an imitation fingerprint from the output of the pattern classifier.

The step of calculating the distribution characteristic of the method includes extracting a feature of an image of a specific region in a plurality of absorbed images and scattering images, and calculating distribution characteristics such as mean and variance of the extracted features.

In addition, the imitation fingerprint determining step extracts a feature of an image of a specific region in the first image and the second image, uses the extracted features as an input to the pattern classifier, and obtains an output of the pattern classifier. And determining whether an object in contact with the fingerprint image acquisition device is a biometric fingerprint or a fake fingerprint.

In the above method, the feature of the image may be, for example, an average value of gray levels, but is not limited thereto. Extracting various other features may be easily understood by those skilled in the image processing art.

In addition, the distribution characteristics of the extracted features may be expressed in a two-dimensional form, such as average, variance, but is not limited thereto, and may be represented in various forms according to the characteristics of the pattern classifier to be used. Ramen is easy to see.

According to the present invention, the additional configuration of hardware and software is small and the counterfeit fingerprint discrimination algorithm is simple, and it is possible to accurately determine whether the object in contact with the fingerprint input window is a bio fingerprint or a counterfeit fingerprint.

Hereinafter, with reference to the drawings will be described an embodiment of a method and apparatus for determining a biometric fingerprint and a fake fingerprint in the optical fingerprint image acquisition device according to the present invention. The following description mainly describes embodiments of the method, and it will be apparent to those skilled in the art that an embodiment of the apparatus operated by the method according to the following description will be appreciated.

7 is a flowchart of a method for discriminating a biometric fingerprint and a fake fingerprint in an optical fingerprint image acquisition device according to the present invention.

First, a fingerprint image (700) acquired by an absorption method for a plurality of biological fingerprints (hereinafter, abbreviated as' absorption image 'for convenience, but for convenience) and a fingerprint image (700') obtained by a scattering method (hereinafter, accurate Although not defined, for convenience, the characteristics of each biofingerprint and imitation fingerprint are extracted from the scattering image to obtain distribution characteristics (e.g., average, variance, etc.) in the feature space. (710, 710 '). This is explained in more detail.

Feature values for each image can be extracted in a variety of ways. FIG. 8 illustrates one method of acquiring an image from a biofingerprint and extracting feature values for the image. First, absorbed images and scattered images are obtained for a plurality of biological fingerprints (810, 810 ′), and gray level average values of predetermined regions are calculated for each image (820, 820 ′). When the gray level average value of the predetermined region for the absorbed image is called the first feature value, and the gray level average value of the predetermined region for the scattered image is called the second feature value, this corresponds to a point in the two-dimensional feature space. If the process is repeated for all bio fingerprints 800, a unique distribution map may be obtained instead of a single point in a feature space corresponding to a single bio fingerprint (830). Although FIG. 8 illustrates only the biofingerprint 800, the above description is also the same for a plurality of fake fingerprints. That is, if the above process is performed on a plurality of imitation fingerprints, a unique distribution on the feature space corresponding to the imitation fingerprints (distribution map of gray level average values for each imitation fingerprint image) can be obtained.

Fig. 9 illustrates the distribution in the feature space of a plurality of bio fingerprints and fake fingerprints calculated in this way. In FIG. 9, the distribution of feature values (gray level average values in FIG. 9) extracted from the absorbed image is mainly formed in the upper left portion, and the distribution of feature values (gray level average value in FIG. 9) extracted from the scattering image is mainly in the lower right portion. It can be seen that it is formed. Although it will be described in advance, an unknown object input to the fingerprint input window (not shown) of the fingerprint acquisition device through the feature value distribution diagram in the feature space as shown in FIG. 9 ("unknown fingerprint" which does not know whether the object is a biometric fingerprint or a fake fingerprint). If the characteristic value of the absorption image and the characteristic value of the scattering image are located at the point A in FIG. 9, the present invention determines whether the unknown fingerprint located at the point A is a bio fingerprint or a fake fingerprint. The process involves designing a pattern classifier. The pattern classifier will be described later.

In FIG. 8, as another embodiment of extracting feature values for each image, binarization is performed on a predetermined region without directly calculating a gray level average value of a predetermined region of each image (840 and 840 '). After that, it is also possible to calculate the gray level average value for the predetermined area.

In addition to the examples shown above, there are a variety of methods for extracting feature values for each image, and it is not limited to extracting a suitable feature in which each class can be easily distinguished in the feature space.

Returning to FIG. 7 again.

In this way, if the distribution characteristics of a plurality of bio fingerprints and imitation fingerprints are obtained, a pattern classifier can be designed using the distribution characteristics (720). The method of designing the booster pattern classifier by its distribution characteristics is theoretically very well organized. Typical examples include methods using statistical features, methods using structural features, and methods using neural networks. For example, "Pattern Recognition: Statistical, Structure and Neural Approach, R. Schalkoff, John Wiley and Sons, 1992." It is well known in the prior art, such as ", and will be apparent to those skilled in the art, detailed description thereof will be omitted.

After designing the pattern classifier, if an unknown fingerprint comes into contact with a predetermined fingerprint input window (not shown) for fingerprint authentication (730), the fingerprint classifier may absorb the unknown fingerprint from the unknown fingerprint in contact with the fingerprint input window of the fingerprint image acquisition device. The l-image is acquired (740), and the first feature value is extracted (750) from the first image. Similarly, the second image is acquired by the scattering method (740 ') and the second feature value is extracted from the second image (750'). In the conventional general fingerprint acquisition device obtained a fingerprint image using only one of the absorption method or scattering method, in the present invention, the absorption image (first image) and scattering image (the first sensor) in the same sensor to identify the imitation fingerprint 2 images).

Next, the first feature value and the second feature value are input to the pattern classifier (760), and it is determined whether the unknown fingerprint is a biometric fingerprint or an imitation fingerprint from the result value from the pattern classifier (770). As described above, it is a step of determining whether the absorbed image and the scattered image feature value of the unknown fingerprint are located at the point A of FIG. 9 by using the pattern classifier designed above.

Although the preferred embodiments of the present invention have been described above with reference to the drawings, the technical scope of the present invention is not limited to the above embodiments but will be determined by the appended claims.

1 is a block diagram of a general absorption type fingerprint image acquisition device.

2 is an explanatory diagram illustrating a principle of an absorption type fingerprint image acquisition device.

3 is a block diagram of a general scattering fingerprint image acquisition device.

4 is an explanatory diagram illustrating a principle of a scattering fingerprint image acquisition device;

5 is an image photograph obtained by the biometric fingerprint in each of the absorption method and the scattering method.

6 is an image photograph obtained by the imitation fingerprint and scattering method respectively.

7 is a flowchart of a method for identifying a fingerprint according to the present invention.

FIG. 8 is an exemplary process diagram of step 710 of FIG. 7; FIG.

9 is an exemplary diagram illustrating a distribution characteristic of a gray level average value calculated in FIG. 8.

FIG. 10 is an exemplary diagram of regions extracted from respective binarized fingerprint images of FIG. 8; FIG.

Claims (8)

Extracting feature values from absorbed images and scattered images obtained from a plurality of biological fingerprints and imitation fingerprints, and obtaining distribution characteristics in a feature space for each; Configuring a pattern classifier for classifying each distribution pattern from distribution characteristics of the biofingerprint and the imitation fingerprint; Acquiring a first image by an absorption method and a second image by a scattering method from an unknown fingerprint in contact with the fingerprint image acquisition device for fingerprint authentication; Extracting feature values from the first image and the second image, and inputting the feature values to the pattern classifier to obtain an output of the pattern classifier; And determining, from the output of the pattern classifier, whether the unknown fingerprint in contact with the fingerprint image acquisition device is a biometric fingerprint or a fake fingerprint. The method of claim 1, wherein the extracting of the feature values comprises extracting feature values of an image of a predetermined area in a plurality of absorbed images and scattered images. The method of claim 1, wherein the feature value is an average value or a variance value of gray levels of the fingerprint image. The method of claim 1, wherein the obtaining of the distribution characteristic on the feature space comprises: Comprising the steps of extracting the features of the image of a particular area in the plurality of absorption images and scattering image, and calculating the distribution characteristics such as the average or variance of the extracted features, bio fingerprint and imitation fingerprint in the optical fingerprint image acquisition device How to determine Means for extracting feature values from absorbed images and scattered images obtained from a plurality of biofingerprints and imitation fingerprints to obtain distribution characteristics in a feature space for each; Pattern classification means for classifying each distribution pattern from the distribution characteristics of the biometric fingerprint and the imitation fingerprint; In order to perform fingerprint authentication, a first image is acquired by an absorption method from an unknown fingerprint contacted with a fingerprint image acquisition device, a second image is obtained by a scattering method, and feature values are extracted from the acquired first and second images. And means for inputting the feature value to the pattern classifier to determine whether an unknown fingerprint in contact with the fingerprint image acquisition device is a biometric fingerprint or a fake fingerprint by the result value output from the pattern classifier. A device for determining a biometric fingerprint and a fake fingerprint in the device. 6. The apparatus of claim 5, wherein the feature value comprises means for extracting feature values of an image of a predetermined region in a plurality of absorbed images and scattered images. 6. The apparatus of claim 5, wherein the feature value is an average value or a variance value of gray levels of the fingerprint image. The method of claim 5, wherein the means for obtaining the distribution characteristic on the feature space includes: Biometric fingerprints and imitation fingerprints in the optical fingerprint image acquisition device, characterized in that to extract the characteristics of the image of a particular area in the plurality of absorption images and scattering image, and to calculate the distribution characteristics such as the average or variance of the extracted features Device to determine

Images