JP2003271961A - Method and device for authenticating fingerprint - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To improve authentication precision for a user having a fingerprint in which many longitudinal stripes exist. <P>SOLUTION: In this fingerprint authenticating method wherein an authentication side fingerprint image data collected by a sensor for reading a fingerprint image in every one line is collated with a registration side fingerprint image data, and wherein the subject is identified based on a judge-determination result therein, a feature of the stripe in each line is detected based on a concentration change of a picture element in each line in the authentication side fingerprint image data, when an integrated evaluation value used for evaluating integrally a conformity degree between the authentication side fingerprint image data and the registration side fingerprint image data is calculated, the integrated evaluation value is calculated using a value providing a feature of the detected stripe, and the subject is identified based on the integrated evaluation value. The purpose of the present invention is solved by the fingerprint authenticating method. <P>COPYRIGHT: (C)2003,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a fingerprint authentication method and device in a fingerprint authentication device using a line sensor, and more particularly to a fingerprint authentication method and device capable of reducing the false rejection rate.

[0002]

2. Description of the Related Art The following method has been considered as one method of fingerprint authentication in a fingerprint authentication device using a line sensor. In this method, a predetermined line in the fingerprint data registered in advance at the time of the authentication operation by the user,
A line evaluation value representing the degree of matching between the predetermined lines of the fingerprint data read at the time of authentication is obtained, and the obtained line evaluation value is evaluated over the entire fingerprint image so that a comprehensive evaluation value can be obtained. There is. Then, by performing authentication using this comprehensive evaluation value at the time of authentication, the CPU
It is possible to provide a fingerprint authentication device that minimizes the processing load of.

[0003]

As described above, the line evaluation value obtained by the fingerprint authentication device becomes higher as the number of pixels between lines to be compared increases. However, it has been confirmed at the time of authentication that there are many cases where the pixel values do not match at the boundary of the fingerprint stripes and that there are many cases where the pixel values match inside the stripe.

The occurrence of such a phenomenon causes the following problems.

In the case of a fingerprint having many vertical stripes, the overall evaluation value is distributed low.

In the case of a fingerprint having many horizontal stripes, the comprehensive evaluation value is distributed higher. In the case of, in order to prevent the false acceptance rate (probability that the person to be authenticated is OK even if the person to be authenticated is not the person), the threshold for determining the person may be set to a high value. In the case of a fingerprint with a large number of fingerprints, the overall evaluation value tends to be distributed lower, resulting in a higher rejection rate of the person. That is, in the case of a user having a fingerprint with many vertical stripes, the authentication accuracy at the time of authentication is deteriorated, which makes the user inconvenient.

The present invention has been made in view of the above problems. An object of the present invention is to provide a fingerprint authentication method and apparatus capable of improving the authentication accuracy of a user having a fingerprint with many vertical stripes. Is to provide.

[0008]

In order to solve the above-mentioned problems, according to the present invention, as described in claim 1, authentication side fingerprint image data collected by a line sensor for reading a fingerprint image line by line is obtained. In the fingerprint authentication method that matches the registered fingerprint image data and identifies the person based on the determination result,
When calculating a comprehensive evaluation value used for comprehensively evaluating the degree of coincidence between the fingerprint image data on the authentication side and the fingerprint image data on the registered side, from the change in shading of pixels of each line of the fingerprint image data on the authentication side, The feature of the present invention is that the stripe feature in each line is detected, the comprehensive evaluation value is calculated using a value giving the detected stripe feature, and the person is identified by this comprehensive evaluation value.

According to such a fingerprint authentication method, the characteristic of the stripe in the fingerprint image on the authentication side is detected from the change in the density of the pixels of each line, and the correction value given according to the detection result is used to calculate the total evaluation value. Applied when. That is, even for a fingerprint with many vertical stripes, the correction amount that improves the overall evaluation value is determined according to the characteristics of the vertical stripes, so that the authentication accuracy can be improved and the usability for the user can be improved. it can.

In order to solve the above-mentioned problems, a second aspect of the present invention is provided.
As described in, in the fingerprint authentication device that collates the fingerprint image data on the authentication side collected by the line sensor that reads the fingerprint image for each line with the fingerprint image data on the registration side, and identifies the person based on the determination result, When calculating a comprehensive evaluation value used for comprehensively evaluating the degree of coincidence between the fingerprint image data on the authentication side and the fingerprint image data on the registered side, from the change in shading of pixels of each line of the fingerprint image data on the authentication side, A stripe feature detecting means for detecting the stripe feature in each line, a correction value calculating means for obtaining a value giving the detected stripe feature, and a comprehensive evaluation for calculating the comprehensive evaluation value using the correction value It is characterized by having a value calculation means and identifying the person by this comprehensive evaluation value.

The features of the stripes in the fingerprint image according to the present invention are as follows:
For example, as described in claim 3, in the fingerprint authentication device, the stripe feature detecting means detects the stripe feature in each line based on the number of black and white change points of the density of the pixel in each line. It has a feature.

In such a fingerprint authentication device, the feature of stripes can be detected from the black-and-white change in the density of the pixels on each line.

From the same viewpoint as described above, the present invention provides claim 4.
As described in the above, in the fingerprint authentication device, the stripe feature detection means is based on the average frequency of each line,
The feature is that the feature of the stripe in each line is detected.

In such a fingerprint authentication device, the feature of stripes can be detected from the average frequency of each line.

As for the correction value according to the present invention, as described in claim 5, in the fingerprint authentication apparatus, the correction value calculation means is the stripe feature detection means, and the black and white change of the density of the pixel of each line. The feature is that the correction value is calculated using the features of the stripes detected based on the score or the average frequency of each line.

In such a fingerprint authentication device, for example, after binarizing the pixels of each line, the number of times white and black change is counted, and the larger the number of black and white alternations, the larger the total evaluation value. to correct. As the correction value, an appropriate value is given to the horizontal line in which vertical stripes are detected in this way so that the comprehensive evaluation value does not deteriorate. On the contrary, no correction is performed on the line in which no vertical stripe is detected. Therefore, according to the fingerprint authentication device of the present invention, even if the fingerprint has many vertical stripes, it becomes possible to perform the user authentication without deteriorating the comprehensive evaluation value, so that the user rejection rate can be improved. Further, even if the average frequency of lines is used for detecting vertical stripes, the same effect as above can be expected.

The comprehensive evaluation value relating to the present invention is, for example, as described in claim 6, in the fingerprint authentication device, the comprehensive evaluation value calculating means is a correction value obtained by the correction value calculating means. A fingerprint authentication device, characterized in that a comprehensive evaluation value is calculated according to the following formula.

[0018]

[Equation 3] E ': Total evaluation value Ai: Line evaluation value α, β: Constant Ci: Number of black and white changes in density of each line + 1 (1 if white or black only) n: Number of lines As for the comprehensive evaluation value, as described in claim 6, in the fingerprint authentication method, the comprehensive evaluation value calculating means uses the correction value obtained by the correction value calculating means, and the comprehensive evaluation value is calculated according to the following formula. A fingerprint authentication device characterized by calculating.

[0019]

[Equation 4] E ′: Overall evaluation value E: Overall evaluation value without correction α, β: Constant C: Average frequency of stripes on each line Further, from the viewpoint that it is also possible to correct the fingerprint image data on the registration side. According to the present invention, as described in claim 8, the authentication side fingerprint image data collected by the line sensor that reads the fingerprint image for each line is collated with the registration side fingerprint image data, and based on the determination result, the person In the fingerprint authentication device for identifying, when calculating an overall evaluation value used for comprehensively evaluating the degree of coincidence between the authentication side fingerprint image data and the registration side fingerprint image data, the registration side fingerprint image data Stripe feature detection means for detecting the stripe feature in each line from the grayscale change of pixels in each line, correction value calculation means for obtaining a value giving the detected stripe feature, and using the correction value, Comprehensive And a total evaluation value calculating means for calculating a value, and wherein identifying a person by the overall evaluation value.

[0020]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a diagram showing a configuration example of a fingerprint authentication system to which the fingerprint authentication method according to the present invention is applied. This fingerprint authentication system includes a fingerprint authentication device 1 including a line sensor that scans and reads a user's fingerprint image line by line.
00 and a host computer 200. The fingerprint data registration device 100 stores (= registers) the fingerprint data read and collected by the line sensor 101 for each line in a predetermined memory. The fingerprint data stored in the memory is sent to the host computer 200 through the communication line, and the user can select the fingerprint data to be registered according to the instruction on the host computer 200. In addition, in the fingerprint authentication system of this example, the fingerprint data registered at the time of authentication is collated, and only the user who is regarded as legitimate by the collation is permitted to use the host computer 200.

FIG. 2 is a block diagram showing a schematic configuration of the fingerprint authentication device 100 shown in FIG. A fingerprint data registration device 100 shown in the figure includes a line sensor 101 that scans and reads a user's fingerprint image line by line, and an analog-to-digital converter A / D (Analog) that converts analog line data output by the line sensor 101. / Digital conver
tor) 102 and a control unit (CPU) 103 for extracting the contour of the finger of the fingerprint image read by the line sensor 101,
It is composed of a memory unit 120. The memory unit 120 is
An SRAM (static random access memory) that instantly stores fingerprint data read by the line sensor 101 when a user places a finger on the line sensor 101 and slides the finger.
y) 104 and a flash memory (FROM: flash Rea) having a non-volatile memory area for storing programs such as a fingerprint authentication algorithm and fingerprint data for fingerprint matching.
d Only Memory) 105.

The fingerprint authentication device 100 configured as described above
With reference to FIG. 3, an example of a processing procedure for personal authentication according to the fingerprint authentication method according to the present invention will be described below. FIG. 3 is a flowchart showing an example of a processing procedure for personal authentication according to the fingerprint authentication method according to the present invention.

One fingerprint image data read by the line sensor 101 at the time of fingerprint authentication is acquired, for example, as a gray scale image of 256 lines in the vertical direction, 256 pixels in the horizontal direction, and 0 to 255 in 256 steps (S1). That is, 256 lines of fingerprint data are collected as fingerprint data.
The control unit 103 performs preprocessing (S2) using the fingerprint data thus collected. This pretreatment (S2)
Is a process for extracting the contour of the finger, and more specifically,
This is performed according to the procedure shown in FIG. FIG. 4 is a flowchart showing an example of the preprocessing shown in FIG.

(Description of Pre-Processing) This pre-processing is performed in the following three steps.

Normalization (S1) Contour extraction (S2) Binarization (S3) The line data for each scan read by the line sensor 101 is converted into digital information by the A / D 102, and then SR
It is stored in the AM 104. The control unit 103 uses the SRAM 1
Take out the fingerprint data for one image stored in 04,
Normalization processing (S11) for removing density unevenness and noise is performed. For example, a simple average value (here, abbreviated as (A)) of pixel values (= density value (gradation value)) of 11 × 11 pixels around the target pixel (= pixel to be binarized) is calculated. From this (A), the difference (see FIG. 5) from the average density value (here, abbreviated as (B)) of the left and right three pixels adjacent to the target pixel is obtained. As a result, the LED of the line sensor 101
It is possible to reduce the influence of uneven illuminance and uneven density of a fingerprint image caused by suppressing the finger on the line sensor.

When the image normalization processing is completed as described above, the control unit 103 proceeds to the next step (S12) and performs processing for determining the position of the contour of the finger. This process is performed by, for example, grasping at which pixel the intensity change has occurred for each line of interest by detecting the vertices (extreme values) by the first-order differentiation operation, and determining the pixel position obtained by the extreme value search as the contour position. Make it a candidate. However, it is estimated that the contour position candidate pixels obtained at this stage have the ridges or valleys of the fingerprint groove, and are considered to contain noise. For this reason, a position that is significantly different from the true position may be determined to be a mountain or a valley. Therefore, the pixel position of the contour candidate is compared with the position of the predetermined reference line to correct the reference line. As a result, even if peaks and valleys are found at positions that are greatly different by about 1 to 2 points, the reference line is not largely corrected, so that the influence of noise can be reduced. Also, by correcting the reference line as described above,
The finger and the outside of the finger are discriminated, and the data on the outside of the finger that is not valid for fingerprint authentication is discarded. When the contour position of the finger is determined as described above, the control unit 103 performs the final step, that is, the binarization process (a monochrome 256-tone grayscale image (grayscale image) after contour extraction is performed in white and black 2). The process of converting to a gradation image) is performed (S13), and the processed image is used as the image data for fingerprint authentication from the FROM 105.
Output to and accumulate.

Here, returning to FIG. 3, when the preprocessing (S2 in FIG. 3) is finished as described above, the control unit 103 determines the line data (a line data indicating a predetermined position) from the line data of the preprocessed image. Hereinafter, line data A is abbreviated), and a predetermined area in the image registered in advance is searched to find a line (= reference line) that best matches the line data A. In this example, this search processing is referred to as reference line search processing. Also, in order to find the line on the registration side that best matches the line data A, an evaluation function called "line evaluation value" is introduced. What is line evaluation value 2
It is a value indicating the degree of coincidence of two lines. That is, the registration side line that has output the maximum line evaluation value for the line data A is adopted as the reference line. After obtaining the reference line, the control unit 103 sequentially obtains the maximum line evaluation value Ai for each line i with this reference line as the start position, and corrects the line evaluation value Ai to make a comprehensive evaluation value. Find E '. The correction value used when obtaining the comprehensive evaluation value E ′ is given by, for example, a value representing the number of times of black and white replacement for each line i, and the comprehensive evaluation value E ′ is calculated by the following formula.

[0029]

[Equation 5] E ': Total evaluation value Ai: Line evaluation value α, β: Constant Ci: Correction amount (number of black and white change points of each line) n: Number of lines In the above formula, α and β are predetermined constants, (Ci-1)
Ci in the +1 term is the above-mentioned correction value. This correction value is determined as follows, for example.

(Example of determination of correction value) When the control unit 103 obtains a fingerprint image having the number of horizontal pixels (N = 256) and the number of vertical lines (M = 256) after the pre-processing, each line of each N is obtained. The pixel value of is checked and the number of black and white changes is stored in a predetermined storage area (S3). For example, when N = 9, if there is a black-and-white alternation of each vertical line, it is determined that the horizontal line (N = 9) has a vertical stripe, and the black-and-white alternation is determined according to the number of alternations. Determine the correction value. In other words, this correction value serves as a weight for preventing the overall evaluation value from deteriorating. On the other hand, when there is no black-and-white alternation number in each vertical line of the horizontal line (N = 9) (only one of white and black), the horizontal line (N = 9)
It is determined that there is no vertical stripe in the image and no correction is performed (Ci is the minimum value, that is, correction value = 0). Moreover, the control unit 103 can select the correction amount according to the amount of Ci by appropriately determining the constants α and β. As a result, even if the fingerprint image contains many vertical stripes, a correction value is determined according to the black and white density of each vertical line, and an appropriate correction is made for the line determined to be vertical stripes. Therefore, it is possible to avoid deterioration of the authentication accuracy.

The control unit 103 sets the correction value determined as described above to the maximum line evaluation value A for each line i.
i is multiplied, and the sum is divided by the number of lines to obtain a total evaluation value E '(see above) (S4). Then, an authentication process (S5) for identifying the person is performed using the total evaluation value.

Next, the results of experiments (see Tables 1 and 2) in which a comprehensive evaluation value is obtained and personal authentication is performed according to the above formula are shown. In this experiment, β, which is a constant, is fixed to 1024,
This is the result of changing α. Also, the comprehensive evaluation value E '
Is the result of brute force calculation of 780 data pieces of 10 Japanese index fingers, middle finger, ring finger, and total of 6 fingers of 13 Japanese men. In addition, the percentage in the table is the rejection rate of the person when the threshold is adjusted so that the acceptance rate of others is 1/10000 (even if the authenticated person is the person, it is not recognized as the person and becomes NG. Probability). Table 1 shows the false rejection rate when there is no correction and Table 2 shows the rejection rate when there is correction.

[0033]

[Table 1]

[0034]

[Table 2] As the above experimental results show, the line evaluation value A for each line
If i is not corrected (see Table 1), the false rejection rate is 2
It will be 5.8%, but with the correction, the maximum is 19.1%.
The rejection rate is improved until (when α is 32). That is, by performing the fingerprint authentication according to the above algorithm, the rejection rate of the person who registered the fingerprint can be reduced, and the usability for the user can be improved.

In the above embodiment, it is assumed that the line evaluation value of each line i is corrected using the fingerprint data on the authentication side (representing the fingerprint data read by the line sensor 101 at the time of authentication). The present invention is not limited to this, and fingerprint data on the registration side may be employed to correct each line i. Further, both the fingerprint data on the authentication side and the fingerprint data on the registration side may be corrected and averaged. Further, in the above-described embodiment, the case where the line evaluation value for each line i is corrected has been described as an example. However, when calculating the total evaluation value, the total evaluation value itself may be corrected. Absent. In this case, the total evaluation value E
′ Is calculated, for example, by the following equation.

[0036]

[Equation 6] E ': Comprehensive evaluation value E: Comprehensive evaluation value without correction α, β: Constant C: Fingerprint stripe frequency (average frequency) The above E represents the comprehensive evaluation value without correction, and C is the authentication side. It is the average frequency of the stripes of fingerprint data (the sum of the lengths of continuous black and white divided by the sum of the number of alternations). As an experiment, the average frequency of the stripes of the fingerprint data on the authentication side was measured, the measured value was substituted into the above formula, and a comprehensive evaluation value E'given a given constant for α and β was calculated. The rate improved to 13.9%. The above equation is
This is merely an example, and various modifications may be made without departing from the spirit of the present invention. Further, C does not deviate from the gist of the present invention even if it is the average number of times of black and white replacement, not the average frequency.

In the above example, the arithmetic processing function of the control unit 103 corresponds to the stripe feature detecting means, the correction value calculating means, and the total evaluation value calculating means.

[0038]

As is apparent from the above description, according to the present invention, vertical stripes are detected by checking the change in density of each line of the fingerprint data (black and white density of pixels), and only the amount of detection is detected. Since the matching correction amount is weighted with respect to the line evaluation value, even if the fingerprint has many vertical stripes, it is possible to perform personal authentication in a state where the overall evaluation value is improved, and as a result,
The rejection rate can be reduced. Therefore, it is possible to improve usability for a user who has many vertical stripe fingerprints.

[Brief description of drawings]

FIG. 1 is a diagram showing a configuration example of a fingerprint authentication system to which a fingerprint authentication method according to the present invention is applied.

FIG. 2 is a block diagram showing a schematic configuration of the fingerprint authentication device shown in FIG.

FIG. 3 is a flowchart showing an example of a processing procedure for personal authentication according to the fingerprint authentication method according to the present invention.

FIG. 4 is a flowchart showing an example of the preprocessing shown in FIG.

[Explanation of symbols]

100 Fingerprint data registration device 101 line sensor 102 A / D 103 control unit (CPU) 104 SRAM 105 FROM (flash ROM) 120 memory 200 host computer 300 communication lines

Claims (8)

[Claims] 1. A fingerprint authentication method in which fingerprint image data on the authentication side collected by a line sensor that reads a fingerprint image for each line is collated with fingerprint image data on the registration side, and the person is identified based on the determination result. When calculating a comprehensive evaluation value used to comprehensively evaluate the degree of coincidence between the fingerprint image data on the side of registration and the fingerprint image data on the side of registration, from the change in the shade of the pixels of each line of the fingerprint image data on the authentication side, respectively. The fingerprint authentication method characterized in that the feature of the stripe in the line is detected, the comprehensive evaluation value is calculated using a value giving the detected feature of the stripe, and the person is identified by the comprehensive evaluation value. 2. A fingerprint authentication device for collating authentication-side fingerprint image data collected by a line sensor that reads a fingerprint image for each line with registration-side fingerprint image data, and for identifying the person based on the determination result. When calculating a comprehensive evaluation value used to comprehensively evaluate the degree of coincidence between the fingerprint image data on the side of registration and the fingerprint image data on the side of registration, from the change in the shade of the pixels of each line of the fingerprint image data on the authentication side, respectively. Stripe feature detecting means for detecting the stripe feature in the line, correction value calculating means for obtaining a value giving the detected stripe feature, and total evaluation value calculation for calculating the total evaluation value using the correction value. And a means for identifying the person based on the comprehensive evaluation value. 3. The fingerprint authentication device according to claim 2, wherein the stripe feature detecting means detects the stripe feature in each line based on the number of black and white change points of the density of the pixel in each line. Fingerprint authentication device. 4. The fingerprint authentication device according to claim 2, wherein the stripe feature detection means detects a stripe feature in each line based on an average frequency of each line. 5. The fingerprint authentication apparatus according to claim 3 or 4, wherein the correction value calculation means detects the stripe characteristic detection means based on the number of black and white change points of the density of pixels of each line or the average frequency of each line. A fingerprint authentication device characterized in that a correction value is calculated using the feature of the striped pattern. 6. The fingerprint authentication apparatus according to claim 2 or 5, wherein the total evaluation value calculation means calculates the total evaluation value according to the following equation using the correction value obtained by the correction value calculation means. A fingerprint authentication device characterized by. [Equation 1] E ′: Overall evaluation value Ai: Line evaluation value α, β: Constant (Ci−1) +1: Number of black and white change points of density of each line n: Fingerprint authentication method characterized by calculation according to the number of lines. 7. The fingerprint authentication method according to claim 2 or 5, wherein the comprehensive evaluation value calculation means calculates the comprehensive evaluation value according to the following equation using the correction value obtained by the correction value calculation means. A fingerprint authentication device characterized by. [Equation 2] E ': Total evaluation value E: Total evaluation value without correction α, β: Constant C: Average frequency of stripes on each line 8. A fingerprint authentication device for collating the fingerprint image data on the authentication side collected by a line sensor for reading a fingerprint image line by line with the fingerprint image data on the registration side to identify the person based on the judgment result. When calculating an overall evaluation value used for comprehensively evaluating the degree of coincidence between the fingerprint image data on the registration side and the fingerprint image data on the registration side, from the change in shading of pixels in each line of the fingerprint image data on the registration side, respectively. Stripe feature detecting means for detecting the stripe feature in the line, correction value calculating means for obtaining a value giving the detected stripe feature, and total evaluation value calculation for calculating the total evaluation value using the correction value. And a means for identifying the person based on the comprehensive evaluation value.