JPH06274602A - Fingerprint registering/collating device - Google Patents

Abstract

PURPOSE:To extract a relevant template from a dictionary without any ID number and to collate it by providing a fingerprint sensor, dictionary ID prepara tion and template dictionary. CONSTITUTION:Dictionary ID preparation 14 prepares a dictionary ID by encoding a fingerprint read by a fingerprint sensor 1 based on the number of respective feature points or the kind (either an end point or a branch point) in plural divided areas. Corresponding to this dictionary ID, the fingerprint (template) is registered on a template dictionary 9. At the time of collation, the dictionary ID preparation 14 prepares the dictionary ID by encoding the fingerprint read by the fingerprint sensor 1 based on the number or kind of respective feature points in the plural divided areas. Based on this dictionary ID, the relevant template is extracted from the template dictionary 9, and only this extracted template is collated. Thus, fingerprint collation is enabled without inputting any dictionary ID.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a fingerprint registration / collation device for registering and collating fingerprints.

In various fields, security management in access (from opening and closing doors to workstation entry) is a necessity. The personal identification number of the lock system or computer system is an important factor in security management. However, they run the risk of being copied, stolen or forgotten. Therefore, there is fingerprint collation as personal identification using a part of the human body, and a technique for registering this fingerprint and simply collating is required.

[0003]

2. Description of the Related Art Conventionally, as shown in FIG. 12, a fingerprint collation apparatus registers a personal fingerprint template in advance in a fingerprint DB 22 in association with an ID number. Then, when the ID number is input and the fingerprint of the finger is read and input, the collation processing 23 extracts the template from the fingerprint DB 22 based on the ID number, collates with the input fingerprint, and determines OK or NG. Do. The configuration of FIG. 12 will be briefly described below.

FIG. 12 is an explanatory view of the prior art. Figure 1
In FIG. 2, the fingerprint 21 is a fingerprint pattern read from the user's finger.

The fingerprint DB 22 is a template in which the pattern of the fingerprint read from the user's finger is thinned and feature points are extracted and registered in association with the ID number.

The collation processing 23 collates the template extracted from the fingerprint DB 22 with the fingerprint pattern read by the user based on the input ID number, and determines whether or not they match.

[0007]

Conventionally, as described above, at the time of collation, the ID number of the user to be collated is input and the template of the user is taken out from the fingerprint DB 22, and at the same time, the user's template is retrieved. Since the pattern of the fingerprint read from the finger is collated to determine whether they match, there is a problem that a troublesome operation of storing and inputting the ID number becomes necessary. This troublesome I
If the input of the D number is stopped, it is necessary to match with all the templates registered in the fingerprint DB 22, and it is not so much when the number of templates to be matched is small, but when there are many templates, the matching time is There is a problem that it takes a lot of time because it is proportional to the number.

The present invention solves these problems by
Feature points are extracted from the fingerprint and coded, the coded value is registered as a dictionary ID in the fingerprint pattern, and the code generated from the fingerprint at the time of matching is used as the dictionary ID to match only the template extracted from the dictionary. The purpose is to retrieve the corresponding template from the dictionary without the ID number and to enable the matching.

[0009]

FIG. 1 is a block diagram showing the principle of the present invention. In FIG. 1, a fingerprint sensor 1 detects a fingerprint.

The dictionary ID creation 14 encodes the read fingerprint based on the number or type (end point, branch point) of each feature point in the divided area.
The template dictionary 9 registers a fingerprint (template) with the coded value as a dictionary ID.

[0011]

According to the present invention, as shown in FIG. 1, dictionary ID creation 1
For the fingerprint read by the fingerprint sensor 1, the dictionary ID is created based on the number or type (end point, branch point) of each feature point in the divided area, and a dictionary ID is created. In addition, the fingerprint (template) is registered in the template dictionary 9.

Further, at the time of collation, the dictionary ID creation 14 encodes the fingerprint read by the fingerprint sensor 1 into the number or type (end point, branch point) of each feature point in the divided area. Create a dictionary ID and use this dictionary I
Based on D, the corresponding template is retrieved from the template dictionary 9 and only the retrieved template is matched.

Further, the region divided into a plurality of regions is a region divided by a predetermined angle around one feature point. Further, the region divided into a plurality of regions is a region concentrically divided by a predetermined distance around one feature point.

Further, the area divided into a plurality of areas is an area divided by a predetermined angle and concentrically by a predetermined distance centering on one feature point. Further, the area divided into a plurality of areas is divided into a predetermined rectangular size.

Further, when encoding is performed based on the number or type (end point, branch point) of each feature point in a plurality of divided regions, a dead region (code (Excluded from the region of conversion) 31 to reduce the code change due to the distortion of the fingerprint pattern.

In addition, when encoding based on the number or type (end point, branch point) of each feature point in the divided region, the feature points near the boundary of the divided region are multiplied by weights. Then, the coding is performed to reduce the code change due to the distortion of the fingerprint pattern.

Therefore, the feature points are extracted from the fingerprint, divided into a plurality of areas and coded, and the coded values are used as the dictionary ID.
The template generated from the fingerprint without the ID number is registered in the template dictionary 9 as the dictionary ID, or the template is extracted from the template dictionary 9 using the encoded value as the dictionary ID and collated.
It is possible to register it in or to retrieve only the corresponding template from the template dictionary 9 and collate it.

[0018]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Next, the construction and operation of an embodiment of the present invention will be sequentially described in detail with reference to FIGS.

FIG. 1 is a block diagram showing the principle of the present invention. Figure 1
In the above, the fingerprint sensor 1 reads a fingerprint in response to the user pressing the finger.

The numeric keypad 2 is used to input a numerical value of the price of a product. The A / D 3 converts an analog signal of a fingerprint read by the fingerprint sensor 1 into a digital signal.

The fingerprint registration / collation device 4 is for registering fingerprints and collating fingerprints, and has 5 to 18
It is composed of The I / F 5 is an interface that outputs the signal of the numeric keypad 2 to the bus.

The frame memory 6 stores fingerprint data read by the fingerprint sensor 1. The binarization circuit 7 binarizes the multivalued fingerprint data stored in the frame memory 6 and stores it in the binarization memory 8.

The binary memory 8 stores binary digitized fingerprint data. The template dictionary 9 is a dictionary in which templates are registered (see FIG. 6). CPU10
Performs various processes according to a program, and includes a dictionary creation function 11, a window setting 12, a dictionary search function 13, a dictionary ID creation 14, a feature point extraction 15, and the like.

The feature point extraction 15 thins the fingerprint data to extract feature points (branch points, end points, etc.). The window setting 12 is a window (window, window of a predetermined size) centered on the feature points extracted by the feature point extraction 15.
(Rectangle) is set (see FIG. 3).

The dictionary ID creation 14 is based on the number or type (end point, branch point) of each feature point in the divided area to create a dictionary ID (FIG. 2, FIG. 2).
It mentions later using FIG. 3).

The dictionary creating function 11 creates a template by using the encoded value as a dictionary ID, and the template dictionary 9
To be registered with. The dictionary search function 13 retrieves a corresponding template from the template dictionary 9 based on the dictionary ID created from the fingerprint data read from the user's finger, and performs matching.

The collation circuit 17 collates the fingerprint, and collates the fingerprint read by the fingerprint sensor 1 with the template extracted from the template dictionary 9 based on the dictionary ID generated from the fingerprint. It is a thing.

The line I / F 18 is for outputting a door lock signal for opening the door, for example, when the result of the collation indicates that the person is identified as a match. Next, FIG.
The operation of registering a fingerprint in the template dictionary 9 will be described in detail according to the order shown in the flowchart of FIG.

FIG. 2 shows a fingerprint registration flowchart of the present invention. In FIG. 2, S1 is fingerprint imprinting. For this, the user presses his / her finger on the fingerprint sensor 1 of FIG.

In step S2, a fingerprint image is detected. This is S1
Then, the image of the imprinted fingerprint is read in response to the user impressing by pressing the finger on the fingerprint sensor 1. S3 is binarized. This binarizes the image of the fingerprint read in S2.

In step S4, the line is thinned. This extracts the center line of the line and thins the image of the fingerprint binarized in S3. At S5, feature points are extracted. This is to extract end points and branch points as feature points from the fingerprint image after thinning in S4 (see FIG. 5).

In step S6, a window is set (binary image). This sets a rectangular window of a predetermined size centered on the feature point extracted in S5 (see FIG. 2). S7 is
Make a position and window dictionary.

In step S8, the alignment window is set. this is,
A window serving as a fingerprint reference is set, for example, the center window in FIG. 3 is set as an alignment window. In S9, the coding area is set. This sets, for example, the circle in FIG. 3 as the coding area as the coding area.

In step S10, partial circles of 45 degrees are set around the alignment window, and the number of characteristic points existing in each partial circle is calculated. This is, for example, as shown in FIG.
With respect to the positioning window (center window), 45 ° portions are set as shown in the figure, and the number of characteristic points existing in each region (regions 1 to 8) is calculated. In the case of (a) of FIG. 3, it is calculated that the number of regions 1 2 3 4 5 6 7 8 feature points is 2 1 0 1 1 0 1 0.

In step S11, the number of the partial circle and the number of characteristic points are encoded to obtain a dictionary number. This is performed by obtaining the number of characteristic points for each area as described above in S10 and using this as the dictionary number (dictionary I
D) and 21011010. Then, using this dictionary number (dictionary ID) as an index, a template is created and registered in the template dictionary 9 as shown in FIG.

As described above, the fingerprint read from the fingerprint sensor 1 is binarized, thinned, and feature points are extracted (end points,
A branch point is extracted), a window of a predetermined size is set around the feature point, the number of feature points existing in each of the plurality of divided regions is determined, and the number of feature points is coded to form a dictionary ID. Then, using the dictionary ID as an index, the coordinates of each window, the image data in the window, and the like are shown in FIG.
It is registered in the template dictionary 9 as shown in. As a result, it becomes possible to automatically create a dictionary ID from fingerprint data to create a template, and further register the template ID 9 with the dictionary ID as an index.

FIG. 3 shows a coding explanatory diagram of the present invention. FIG. 3A shows an example of a region for encoding. this is,
As the area, as shown in the figure, eight fan-shaped areas are set at intervals of 45 degrees centering on the alignment window. Here, area numbers 1 to 8 are assigned to these areas as shown. The small rectangle is a rectangular window having a predetermined size centered on the feature point. Here, seven feature points are surrounded by a window.

FIG. 3B shows an example of coding by the number of feature points in each area. More specifically, the area 1 in FIG.
Since there are two windows, "2" and "7", the number of feature points is "2". Similarly, when the number of feature points is obtained, the following is shown in the figure: number of regions 1 2 3 4 5 6 7 8 feature points = code 2 1 0 1 1 1 1 0 1.

Therefore, the code (dictionary ID) at this time
Becomes 21011010. FIG. 3C shows an example in which each region is coded. More specifically, as shown in the drawing, the relationship between the types of characteristic points and codes is defined as follows: 3: branch point and end point 2: end point only 1: branch point only 0: none. For example, in FIG.
When the area 1 of (a) is obtained, the code becomes “3” because there are both end points and branch points. Similarly, when the other areas are obtained, the area 1 2 3 4 5 6 7 8 code 3 2 0 1 2 0 1 0 is obtained as shown below.

Therefore, the code at this time (dictionary ID)
Becomes 32012010. FIG. 3D shows an example of encoding with the number of characteristic points for each type. More specifically, FIG.
Since the window 2 is the branch point and the window 7 is the end point in the area 1 of (a) of FIG. Similarly, when the other areas are obtained, the following is shown.

Area 1 2 3 4 5 6 7 8 Branching point 1 0 0 1 1 0 0 1 0 0 End point 1 1 1 0 0 1 1 0 0 0 Code 11 01 00 00 10 01 00 10 00 Therefore, the code at this time (dictionary ID) Is 11010
It becomes 01001001000.

FIG. 4 shows an example of areas for encoding according to the present invention. As for the area, as shown in the figure, the area is set to 16 as shown by a double circle for every 45 degrees centering on the alignment window. Here, area numbers 1 to 16 are assigned to these areas as shown.
The small rectangle is a rectangular window having a predetermined size centered on the feature point. here,
Seven feature points are surrounded by a window. The encoding is
This is performed in the same manner as in (b), (c) and (d) of FIG.

FIG. 5 shows an example of types of characteristic points of the present invention.
FIG. 5A shows a state in which a branching point and an end point in the fingerprint pattern are surrounded by a window (rectangle) of a predetermined size. The windows 1 to 7 are as follows: 1: end point 2: bifurcation point 3: bifurcation point 4: end point 5: bifurcation point 6: end point 7: bifurcation point.

FIG. 5B shows a branch point. Here, the intersection of the thick black line segments shown in the figure is the branch point.
FIG. 5C shows the end points. Here, the point that is interrupted in the middle of the thick black line segment shown in the figure is the end point.

FIG. 6 shows an example of the template dictionary of the present invention. Here, the template is composed of dictionary ID personal ID window number X coordinate Y coordinate image data as shown below. Here, the dictionary ID is a coded value as shown in (b), (c), and (d) of FIG. The window numbers are the numbers 1 to 7 of the feature point windows in FIG. The X coordinate and the Y coordinate are the coordinates of the window of the window number (such as the coordinates at the upper left of the window). The image data is the image data in the window.

Next, the operation at the time of fingerprint collation will be described in detail with reference to FIGS. FIG. 7 shows a configuration diagram at the time of matching according to the present invention. In FIG. 7, the template dictionary 9
Is a window number associated with the dictionary ID, as shown in FIG.
A template storing X coordinates, Y coordinates, and image data is registered.

The search processing 19 uses the dictionary I created from the fingerprint.
Based on D (value coded in any of (b), (c), and (d) of FIG. 3), the corresponding template is retrieved from the template dictionary 9.

The collation processing 20 collates (pattern matching) the fingerprint data (coordinates of windows surrounding the feature points, image data) with the template retrieved from the template dictionary 9 and determines whether or not they match. Is to determine.

Next, the operation of the configuration of FIG. 7 (during collation in FIG. 1) will be described in detail in the order shown in the flowchart of FIG. In FIG. 8, in step S21, fingerprint imprinting is performed. In this, the user presses the finger on the fingerprint sensor 1 of FIG.

In step S22, the fingerprint image is detected. This is S
The image of the fingerprint is read in response to the imprinting of the fingerprint at 21. S23 is binarized. This binarizes the image of the fingerprint read in S22.

In step S24, the line is thinned. This is to extract the center line of the thick line and thin the center line of the fingerprint image binarized in S23. In S25, the characteristic points are extracted. This extracts feature points (branch points, end points) from the fingerprint image thinned in S24.

In S26, the center is selected. This selects a feature point window in the vicinity of the center of the image, that is, selects one window in the vicinity of the center of the image of the fingerprint among windows surrounded by a rectangle of a predetermined size centered on the feature point.

In S27, the counting area is set. This sets a window counting area centering on the feature point window selected in S26, for example, eight fan-shaped window counting areas for every 45 degrees in (a) of FIG.

In step S28, characteristic points are counted. This counts the characteristic points for each of the eight areas set in S27, for example, in FIG. In S29, encoding is performed. This encodes the number of windows from the selected feature point window, for example, as shown in the code of FIG. 3 (b).

In step S30, the candidate dictionary is extracted. This is performed by searching the dictionary ID (that is, searching the template dictionary 9 using the value encoded in S29 as the dictionary ID),
The template with is taken out as a candidate.

In step S31, it is determined whether there is a dictionary candidate. YE
In the case of S, collation is performed after S32. On the other hand, NO
In this case, since the template having the same dictionary ID is not registered in the template dictionary 9, S39
Change the center window to another feature point window with
26 and subsequent steps are repeated.

By the above steps S21 to S31 and S39, all the dictionary candidates (templates) having the dictionary ID of the value encoded from the fingerprint are extracted from the template dictionary 9. The extracted dictionary candidates (templates) are collated according to the following procedure.

In S32, since the template having the same dictionary ID is extracted as a candidate from the template dictionary 9 in YES in S31, the window for alignment is scanned. In step S33, the window for alignment in step S32 is scanned, and it is determined whether the window is within a predetermined search range. If YES, it is determined whether it is a temporary alignment point, and if YES, S
35, and if NO, S33 is repeated. On the other hand, S
If NO in 33, reject the candidate dictionary or
Alternatively, when there are a plurality of dictionary candidates, the next dictionary candidate is selected in the processes of S40 and S41 and the process returns to S32, and when all the dictionary candidates are completed, the process is rejected.

In step S35, pattern matching of the matching window is performed. This pattern-matches the image data of the window corresponding to the fingerprint and the image data of the corresponding window of the candidate dictionary (template) extracted from the template dictionary 9.

In step S36, it is determined whether the result of pattern matching is acceptable. In the case of YES, it is determined in S37 whether or not the required number of windows is acceptable, in the case of YES, it is determined that the identity confirmation is OK, and in the case of NO, the process proceeds to S38. On the other hand, if NO in S36, the process returns to S33.

In step S38, it is determined whether all windows have ended. If YES, the process returns to S33. If NO, the process returns to S35. By the above S32 to S38, S40, and S41, the dictionary candidates (templates) extracted from the tempt dictionary 9 are collated with fingerprints, and when a predetermined number or more of windows are pattern-matched and passed, the dictionary ID is It is obtained as the collation result, and the identity verification is completed.

FIG. 9 shows an example of setting a dead area according to the present invention. FIG. 9A shows an example of the dead zone. The dead area 31 is an area that is not counted when encoding. By providing this dead region 31, even if the feature point is shifted a little due to the distortion of the fingerprint pattern, the feature point does not move to the other region beyond the boundary and is not counted. Can be relaxed.

FIG. 9B shows an example of encoding based on the number of feature points in each area of FIG. 9A. Dead zone 31
When compared with the code in FIG.
The code changes from "1" to "0". As a result, it is possible to mitigate the change in code due to the window 3 in the area 4 crossing the boundary and entering the adjacent area due to a slight distortion of the fingerprint pattern.

FIG. 10 shows an example of correction of the set area boundary according to the present invention. This is a correction example that alleviates the instability of the code when a feature point window is applied to the boundary of the region, and calculates the rotation α degrees from the state before rotation and the number of windows when rotated, The weights of 1 or less are multiplied and added to alleviate the code change due to the window near the boundary.

FIG. 11 shows an example of correction for disappearance of feature points according to the present invention. This is the one that alleviates the change in code with respect to the change in the presence or absence of windows. If the input state of the fingerprint pattern is bad, the feature point detection becomes unstable. Since this tendency of instability is particularly large at the edges of the input image, the weight may be reduced in the outer zone to perform encoding.
Here, the inside circle is set as zone 2 and weight w2 is set,
Zone 2 is defined between the inner side and the outer side and has weight w1 (w2
> W1).

[0066]

As described above, according to the present invention,
Feature points are extracted from the fingerprint, divided into a plurality of areas and encoded, and the encoded values are registered in the template dictionary 9 as a dictionary ID, or the template is taken out from the template dictionary 9 using the encoded values as a dictionary ID. ID number (Dictionary ID)
The template generated from the fingerprint can be registered in the template dictionary 9 without any, or only the corresponding template can be retrieved from the template dictionary 9 and collated. As a result, fingerprint collation can be performed without inputting the dictionary ID.

[Brief description of drawings]

FIG. 1 is a principle configuration diagram of the present invention.

FIG. 2 is a fingerprint registration flowchart of the present invention.

FIG. 3 is an explanatory diagram of encoding of the present invention.

FIG. 4 is an example of areas for encoding according to the present invention.

FIG. 5 is an example of types of characteristic points of the present invention.

FIG. 6 is an example of a template dictionary of the present invention.

FIG. 7 is a configuration diagram at the time of matching according to the present invention.

FIG. 8 is a collation flowchart of the present invention.

FIG. 9 is a setting example of a dead area according to the present invention.

FIG. 10 is an example of correction of a set area boundary according to the present invention.

FIG. 11 is an example of correction for disappearance of feature points according to the present invention.

FIG. 12 is an explanatory diagram of a conventional technique.

[Explanation of symbols]

 1: Fingerprint sensor 2: Numeric keypad 3: A / D 4: Fingerprint registration / collation device 5: I / F 6: Frame memory 7: Binarization circuit 8: Binary memory 9: Template dictionary 10: CPU 11: Dictionary creation Function 12: Window setting 13: Dictionary search function 14: Dictionary ID creation 15: Feature point extraction 16: Collation buffer 17: Collation circuit 18: Line I / F 19: Search processing 20: Collation processing 31: Dead area

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁵ Identification number Reference number within the agency FI Technical indication location G06F 15/66 400 8420-5L (72) Inventor Shigeru Yoshida 1015 Kamitadanaka, Nakahara-ku, Kawasaki-shi, Kanagawa Within Fujitsu Limited

Claims (8)

[Claims] 1. A fingerprint registration / collation device for registering a fingerprint, a fingerprint sensor (1) for detecting a fingerprint, and each feature point in a region divided into a plurality of areas of the fingerprint read by the fingerprint sensor (1). It is provided with a dictionary ID creation (14) for encoding based on the number or type (end point, branch point) of the, and a template dictionary (9) for registering a fingerprint with the encoded value as a dictionary ID. Characteristic fingerprint registration / matching device. 2. A fingerprint registration / collation device for collating fingerprints, a fingerprint sensor (1) for detecting a fingerprint, and a fingerprint read by the fingerprint sensor (1). Each feature point in a divided region. And a dictionary ID creation (14) for encoding based on the number or type (end point, branch point) of the template, and the encoded value at the time of matching as a dictionary ID A fingerprint registration / matching device characterized in that the fingerprint is taken out and only the taken-out template is matched. 3. The fingerprint registration / verification device according to claim 1, wherein the area divided into a plurality of areas is an area divided by a predetermined angle around one feature point. 4. The fingerprint registration / verification according to claim 1 or 2, wherein the plurality of divided regions are concentrically divided regions at predetermined distances around one feature point. apparatus. 5. The area divided into the plurality of areas is an area divided by a predetermined angle and concentrically by a predetermined distance around one feature point as a center. Fingerprint registration / verification device. 6. The fingerprint registration / verification device according to claim 1, wherein the plurality of divided regions are regions divided into a predetermined rectangular size. 7. An insensitive area within a predetermined range from the boundary of the divided areas when encoding based on the number or types (end points, branch points) of each characteristic point in the plurality of divided areas. The fingerprint registration / verification device according to any one of claims 1 to 6, characterized in that (31) is set to be excluded from the coding region) so as to alleviate a code change due to distortion of a fingerprint pattern. 8. When encoding based on the number or type (end point, branch point) of each feature point in the region divided into the plurality of regions, weighting is applied to feature points near the boundary of the divided region. 2. A structure in which the coding is performed by multiplication so as to mitigate the code change due to the distortion of the fingerprint pattern.
7. The fingerprint registration / verification device according to claim 6.