JPH07146941A - Collation device - Google Patents

Abstract

PURPOSE:To detect the matching between an input pattern and a reference pattern with high precision. CONSTITUTION:A correlation arithmetic means 30 performs collation arithmetic between the input pattern and reference pattern and supplies a correlative value obtained as a result to a decision means 50, which decides whether or not the input pattern and reference pattern are the same patterns. This decision is made by comparing the correlative value obtained by the correlation arithmetic means 30 with a threshold value determined according to the degree of autocorrelation among plural reference patterns obtained by copying the reference pattern.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a collation device for recognizing patterns by using optical correlation calculation.

[0002]

2. Description of the Related Art Since pattern recognition can be used to identify an individual, for example, a collation device has been conventionally developed which utilizes a fingerprint pattern as personal information and collates an individual from this fingerprint pattern (see Kaihei 1-222357
6 etc.).

Various methods for performing pattern correlation at high speed by optical calculation have been proposed. For example, a method using a hologram or a joint transform correlation (J) method has been proposed.
There is one called TC) ("Fourier transform in optical information processing", Optics 21,6,392-399). This congruential transform correlation method uses a Vander-Lugt optical correlator (“Pattern recognition using an optical correlation system phase filter”, applied physics).
58, 6 (1989)), which has the advantage of not requiring precise alignment of the hologram plate on which the pattern is written.

FIG. 9 is a block diagram showing the configuration of a conventional recognition device for collating fingerprint patterns using the joint transform correlation method. The recognition device shown in the figure uses high-resolution spatial light modulators (SLMs) 200 and 201, and inputs an input pattern (input image) and a reference pattern (reference image) to a CRT2.
02 is drawn at the same time to detect the coincidence of both patterns. The input image and the reference image are written in the SLM 200 and Fourier-transformed to obtain a Fourier pattern.
These patterns are read again by the SLM 201 and subjected to optical Fourier transform, and the correlation signal images of these patterns are given to the photodetector (PD) 203.

If the input image and the reference image match,
In the correlation signal image given to the PD 203, a point of 0th order light corresponding to the intensity of the original pattern appears in the center, and two points of ± 1st order light with high brightness correspond to the positions of the input image and the reference image. It becomes the image that appeared in. The coincidence between the input pattern and the reference pattern can be detected from the points of the ± first-order light in this image.

[0006]

In the above-mentioned conventional apparatus, the ± 1st-order light spot is a signal indicating the correlation strength between the input pattern and the reference pattern. It varies greatly depending on the spatial frequency characteristics. In other words, even if the degree of coincidence of the patterns is 100%, the detected correlation strength greatly differs depending on the type of the patterns.

FIG. 10 shows the result of autocorrelation calculation of the fingerprint pattern of ten fingers using the conventional recognition device. The autocorrelation calculation is a correlation calculation performed between the same patterns as shown in FIG. It can be seen from FIG. 10 that the autocorrelation value has a variation of about ± 20% from a maximum intensity of about 2500 to a minimum intensity of about 1600. This range becomes even larger as the number of samples increases. This is because different fingerprint patterns cause differences in contrast, area, spatial frequency, and the like.

In the conventional collation device, a fixed threshold value is set in advance, and when a correlation degree equal to or higher than this threshold value is obtained, it is determined that the patterns are the same. However, since the autocorrelation value varies from pattern to pattern as described above, this determination may be erroneous.

It is an object of the present invention to solve such a problem and to provide a matching device capable of detecting a match between an input pattern and a reference pattern with high accuracy.

[0010]

In order to solve the above-mentioned problems, the collating apparatus of the present invention performs (a) an autocorrelation calculation of a reference pattern and sets a threshold value corresponding to the autocorrelation strength obtained as a result. A threshold value setting means to be defined, and (b) a correlation calculation means for calculating a correlation between the input pattern and the reference pattern,
(C) A determination unit that determines whether the input pattern and the reference pattern are the same pattern by comparing the correlation value obtained from the correlation calculation unit and the threshold value obtained from the threshold value setting unit.

[0011]

According to the collating device of the present invention, the correlation calculation of the input pattern and the reference pattern is performed by the correlation calculation means, and the correlation value obtained as a result is given to the determination means so that the input pattern and the reference pattern are the same. It is determined whether it is a pattern. This determination is performed by comparing the threshold value determined according to the autocorrelation strength of the reference pattern with the correlation value obtained by the correlation calculation means, and when the correlation value is greater than or equal to the threshold value, , Both patterns are determined to be the same. If the correlation value is less than the threshold value, it is determined that the two patterns are not the same.

[0012]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the accompanying drawings. FIG. 1 is a block diagram showing the configuration of the matching device according to the present embodiment. This collation device is a device that performs personal identification using a fingerprint pattern of a finger. As shown in the figure, the collation device according to the present embodiment detects an input unit 10 that captures a fingerprint input pattern, a memory 20 in which a fingerprint reference pattern and threshold data are registered, and a correlation value between the input pattern and the reference pattern. And a correlation calculation unit 30 that In addition, the threshold setting unit 40 that sets the threshold level based on the reference pattern registered in the memory 20.
And a determination unit 50 that determines whether the input pattern and the reference pattern are the same pattern from the correlation value data output from the correlation calculation unit 30, the correlation calculation unit 30, the threshold value setting unit 40, and the determination unit 50. And a control unit 60 for controlling.

The input section 10 includes a prism 11 and a light emitting diode (LED) 12 for irradiating the prism 11 with light.
A CCD camera 13 for picking up an input pattern of a fingerprint, and C
A lens 14 that forms an input pattern on the CD camera 13
With.

The correlation calculator 30 displays a CRT 301 for displaying an input pattern and a reference pattern side by side, an optical address type spatial light modulator (SLM) 302 for converting the input pattern and the reference pattern into a coherent light image, and a coherent light image. Fourier transform lens 303 to obtain a congruent Fourier transform light image by Fourier transforming Further, an optical address type SLM 304 for converting the combined Fourier transform light image into a coherent light image, a Fourier transform lens 305 for Fourier transforming the coherent light image to obtain a correlation signal image, and a photodetector (PD) for detecting the correlation signal image. And 306. Further, between the SLM 302 and the Fourier transform lens 303, the read laser beam 310 is read to the SLM 302.
SLM3 equipped with half mirrors 311 and 312 for giving
SLM304 between 04 and Fourier transform lens 305
Half mirror 31 for giving a reading laser beam 310 to the
3,314 are provided.

Next, the operation of this embodiment will be described.
The fingerprint pattern is input by using the total reflection input method.
A fingerprint pattern is obtained from the difference in the reflection condition when the collation target applies his / her finger to the prism 11 by irradiating the prism 11 with light from the ED 12.

First, the collation start signal is sent to the control unit 60 by the collation target person pressing a collation start key (not shown).
And the verification process is started. By operating a keyboard (not shown), for example, by inputting a personal identification number, a reference pattern registered in advance in the memory 20 is selected and this reference pattern is sent to the control unit 60.
The reference pattern may be selected by inserting the ID card into the card reader instead of using the keyboard. Next, when the collation target person presses his / her finger against the prism 11, the fingerprint pattern obtained on the surface of the prism 11 is imaged on the CCD camera 13. The CCD camera 13 converts the fingerprint pattern into a video signal and sends it to the control unit 60.

The control unit 60 gives the sent input pattern and reference pattern to the CRT 301 to display them side by side on the screen of the CRT 301 as shown in FIG. 2A (the left half of FIG. Input pattern, reference pattern on the right half). The input pattern and the reference pattern displayed on the CRT 301 are S arranged facing the CRT 301.
Written to LM 302.

The two patterns written in the SLM 302 are read by the read laser beam 310 and are Fourier transformed by the Fourier transform lens 303. And
The combined Fourier transform image as shown in FIG.
Formed on the a-Si layer 304. In addition, SLM3
04 is read by the laser beam 310 and is Fourier transformed by the Fourier transform lens 305. By this Fourier transform, the center zero-order light point as shown in FIG.
A correlation signal image consisting of ± first-order light spots on both sides of the next light is obtained. The light intensity of the −1st-order light spot (or the + 1st-order light spot) in this correlation signal image is detected by the PD 306. The light intensity of the −1 st order light spot corresponds to the correlation between the two patterns. This PD
The detection signal from 306 is sent to the determination unit 50 via the control unit 60.
Threshold value processing is performed using the threshold value data read from the memory 20 to determine whether the input pattern and the reference pattern are the same pattern. The determination result obtained in this way is used, for example, to open and close the door and unlock the computer.

As described above, different reference patterns of fingerprints cause differences in contrast, area, spatial frequency, etc., and the degree of autocorrelation differs for each reference pattern. Therefore, in the present embodiment, the threshold data used when comparing the input pattern and the reference pattern is set for each reference pattern and stored in the memory 20. The threshold value data is set by the threshold value setting unit 40 when the reference target of the fingerprint is registered in advance by the verification target person.

Conventionally, an attempt has been made to determine the matching rate for each individual, as in the document of Japanese Patent Laid-Open No. 63-149776. However, in the case of the present embodiment using the optical correlator, the detected correlation intensity and the coincidence rate do not directly correspond (the exact same pattern, that is, even if the coincidence rate is 100%, the correlation intensity is significantly different). . Therefore, a threshold setting method unique to the optical correlator is required.

In this embodiment, a method of setting a threshold value based on an autocorrelation value and a method of setting a threshold value based on an autocorrelation value and a cross-correlation value are used. The processing procedure of these setting methods will be described with reference to the flowcharts of FIGS. FIG. 3 is a flowchart showing a processing procedure for setting threshold data based on the autocorrelation value. As shown in the figure, first, when the collation target person presses his / her finger against the prism 11 of the input unit 10 to register the fingerprint reference pattern, the CCD camera 13 reads the fingerprint reference pattern (step 70). The read reference pattern is stored in the memory 2 under the control of the control unit 60.
0 is registered (step 71).

Further, the read reference pattern is given to the threshold value setting section to duplicate the reference pattern (step 72). The duplicated reference pattern and the original reference pattern are given to the correlation calculator 30 and the correlation calculation between these patterns is performed (step 73). An intensity signal of the autocorrelation value is obtained by the correlation calculation, and this intensity signal is given to the threshold setting unit 40. Threshold setting section 4
At 0, the threshold value for each reference pattern is determined based on this intensity signal (step 74). The data of the threshold value determined in this way is controlled by the control unit 60.
It is registered in the memory 20 (step 75).

As a method of determining the threshold value, when the collating device of this embodiment is used for a high degree of security such as computer access in a bank or entry / exit of a safe, the threshold value is set to, for example, an autocorrelation value. Set to 80%. If the threshold value is set to such a high level, the probability that a person other than the registrant is erroneously determined to be the registrant becomes almost zero. On the other hand, when the collation device of the present embodiment is used for applications with low security such as entry and exit of general rooms, the threshold value is set to, for example, an autocorrelation value of 40.
Set to%. If the threshold value is set to such a low level, the registrant is rarely judged to be not the registrant, although some erroneous determination is made. In this way, the threshold setting unit 40 sets the threshold according to the security level.

FIG. 4 is a flowchart showing a processing procedure for setting threshold data based on the autocorrelation value and the cross-correlation value. In this example, the cross-correlation value is also used in addition to the auto-correlation value because the fingerprint pattern has the same cross-correlation value as the auto-correlation value. That is, a fingerprint pattern including many features common to other fingerprint patterns outputs a high-intensity cross-correlation. Therefore, an attempt is made to obtain a more accurate threshold value by taking the cross-correlation value into consideration in addition to the auto-correlation value and determining the threshold value.

In this process, first, the prism 1 of the input unit 10 is used in order for the person to be collated to register the reference pattern of the fingerprint.
When the finger is pressed against 1, the CCD camera 13 reads the reference pattern of the fingerprint (step 80). The read reference pattern is registered in the memory 20 under the control of the control unit 60 (step 81).

Further, the read reference pattern is given to the threshold value setting section to duplicate the reference pattern (step 82). The duplicated reference pattern and the original reference pattern are given to the correlation calculator 30, and the correlation calculation is performed between these patterns to detect the autocorrelation value (step 83).

Further, another reference pattern already registered is read from the memory 20 (step 84), and the reference pattern read in step 80 and the other reference pattern are given to the correlation calculating section 30. The correlation calculation unit 30 performs the correlation calculation between these patterns and detects the cross-correlation value (step 85). Both the autocorrelation value detected in step 83 and the cross-correlation value detected in step 85 are given to the threshold setting unit 40 as intensity signals. The threshold value setting unit 40 determines a threshold value for each reference pattern based on these intensity signals (step 86). The threshold data thus determined is registered in the memory 20 under the control of the control unit 60 (step 87).

Next, how to set the threshold value in the threshold value setting section 40 will be described with reference to the experimental results shown in FIGS. 5 to 7 are graphs showing the experimental results of detecting the autocorrelation value and the cross-correlation value of the fingerprint pattern by the correlation calculation. In each graph, reference patterns of different fingerprints (fingerprint A to fingerprint C) are used for the experiment. The horizontal axis of these graphs shows the reference pattern number, and the vertical axis shows the correlation strength when the autocorrelation value is 1.0. First, in the graph of FIG. 5, the reference pattern with the pattern number 0 is the same pattern as the reference pattern of the fingerprint A to be measured, and the correlation value with this pattern is the autocorrelation value. The reference patterns of the fingerprints with the pattern numbers 1 to 29 are fingerprint patterns different from the pattern of the fingerprint A, and the correlation value with these patterns is the cross-correlation value. From this graph, pattern numbers 6 and 18
The cross-correlation value between the second reference pattern and the reference pattern as the measurement target is maximum at 17% of the autocorrelation value. In addition, FIG.
In the graph, the reference pattern with the pattern number 1 is the same pattern as the reference pattern of the fingerprint B to be measured, and the correlation value with this pattern is the autocorrelation value. The correlation value with the reference pattern other than No. 1 is the cross-correlation value. From this graph, the cross-correlation value between the reference patterns of pattern numbers 4, 6, and 29 and the reference pattern of the measurement target is the maximum at 19% of the autocorrelation value. Further, in the graph of FIG. 7, the reference pattern of the pattern number 2 is the same pattern as the reference pattern of the fingerprint C to be measured, and the correlation value with this pattern is the autocorrelation value. Then, the correlation value with the reference pattern other than the second pattern is the cross-correlation value. From this graph, the cross-correlation value between the reference pattern of the pattern number 5 and the reference pattern of the measurement target is maximum at 22% of the autocorrelation value.

The threshold value needs to be determined from the maximum value of this cross-correlation, the auto-correlation value, the size of the detector, noise, the number of registered fingerprints, the safety factor, etc. The maximum value of 17% (in the case of fingerprint A) is the lower limit of the threshold, and the autocorrelation value is the upper limit of the threshold. The level to which the threshold is set within the range from the upper limit to the lower limit is determined according to the purpose. For example, when the collation device of the present embodiment is used for applications with low security such as entry and exit of general rooms, for example, the threshold value is 10% of the upper and lower limits. In the case of fingerprint A, the level is (100% -17%) * 0.1 + 17% = 25.3% as the threshold value. On the other hand, in the case of high security such as bank computer access, for example, 70% of the upper and lower limits are set as the threshold level. Thus, the threshold value may be determined according to the purpose.

In this way, by knowing the upper limit level of the expected correlation value from the autocorrelation value and obtaining the lower limit level of the threshold value from the cross-correlation value, the range (safety factor, security degree) can be determined from this range. By setting the threshold value accordingly, a stable recognition system can be constructed.

In the present embodiment, the threshold value setting section sets the threshold value at the timing when the reference target person registers the reference pattern of the fingerprint in advance, but the threshold value is limited to this timing. Alternatively, for example, it may be performed at the timing when the reference target person inputs the fingerprint input pattern into the input unit 10.

In this embodiment, the fingerprint is used for the person verification, but the fingerprint is not limited to the fingerprint, and a palm pattern or a sole pattern may be used.

Further, in this embodiment, the SLMs 302, 3
Although an optical address type SLM is used as 04, an electric address type SLM may be used.

Instead of the CRT 301, another display device such as a liquid crystal display or a plasma display may be used.

Further, the fingerprint input by the input unit 10 may be performed by the holographic as shown in FIG. 8 other than the prism 11. This sensor
A hologram 101 is formed on a part of a transparent glass flat plate (light guide plate) 100, and laser light is emitted from a laser light source 102 on the opposite side of the other end.
At 3, the fingerprint pattern is detected.

[0036]

As described above in detail, in the collating device of the present invention, the correlation calculation between the input pattern and the reference pattern is performed, and the input pattern and the reference pattern are the same pattern based on the correlation value obtained as a result. Is determined. This determination is performed by comparing a threshold level determined according to the autocorrelation strength between a plurality of reference patterns obtained by duplicating the reference pattern with a correlation value obtained by the correlation calculation.

In this embodiment, since the threshold level is set for each reference pattern according to the autocorrelation strength in this way, it is possible to provide a collation device for performing stable recognition.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a matching device according to an embodiment.

FIG. 2 is a diagram showing an input image, a Fourier image, and a correlation signal image.

FIG. 3 is a flowchart showing a processing procedure of a threshold value setting method.

FIG. 4 is a flowchart showing a processing procedure of a threshold value setting method.

FIG. 5 is a diagram showing an experimental result of detecting an autocorrelation value and a cross-correlation value by a correlation calculation.

FIG. 6 is a diagram showing an experimental result of detecting an autocorrelation value and a cross-correlation value by a correlation calculation.

FIG. 7 is a diagram showing an experimental result of detecting an autocorrelation value and a cross-correlation value by a correlation calculation.

FIG. 8 is a diagram showing an input unit using holography.

FIG. 9 is a block diagram showing a configuration of a conventional recognition device.

FIG. 10 is a diagram showing a result of performing autocorrelation calculation of a fingerprint pattern using a conventional recognition device.

[Explanation of symbols]

10 ... Input part, 11 ... Prism, 12 ... Light emitting diode, 13 ... CCD camera, 14 ... Lens, 20 ... Memory, 30 ... Correlation calculation part, 40 ... Threshold setting part, 50 ...
Judgment part, 60 ... Control part, 100 ... Glass flat plate, 101 ...
Hologram, 102 ... Laser light source, 103 ... Imaging element,
301 ... CRT, 302, 304 ... Spatial light modulator, 30
3, 305 ... Fourier transform lens, 306 ... Photo detector, 310 ... Laser light, 311 to 314 ... Half mirror.

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification number Internal reference number FI Technical indication location 9061-5L G06F 15/70 460 A (72) Inventor Yuji Kobayashi 1126 Nonomachi, Hamamatsu City, Shizuoka Prefecture 1 In Hamamatsu Photonics Co., Ltd. (72) Inventor Tsutomu Hara 1 1126, Nomachi, Hamamatsu City, Shizuoka Prefecture 1 In Hamamatsu Photonics Co., Ltd.

Claims (3)

[Claims] 1. A collation device for determining whether a newly input input pattern and a reference pattern already registered are the same pattern, wherein an autocorrelation operation of the reference pattern is performed, and an autocorrelation strength obtained as a result is calculated. A threshold value setting means for determining a threshold value according to the threshold value; a correlation calculation means for performing a correlation calculation between the input pattern and the reference pattern; a correlation value obtained from the correlation calculation means; and a threshold value setting means. A collation device comprising: a determination unit that compares threshold values and determines whether the input pattern and the reference pattern are the same pattern. 2. The threshold value setting means performs autocorrelation calculation of the reference pattern and cross-correlation calculation of the reference pattern and other registered reference patterns, respectively.
The threshold value is set according to the resulting autocorrelation strength and cross-correlation strength.
The matching device described. 3. The correlation calculation performed by the threshold value setting means and the correlation calculation means is both performed by using a congruential conversion type optical correlator. Matching device.