JPH06195448A - Method and device for collating fingerprint - Google Patents

Abstract

PURPOSE:To reduce erroneous collation ratio by discriminating the feature of a fingerprint pattern in more detail in a moving window fingerprint collating method and its device. CONSTITUTION:In a moving window fingerprint collating device having a binary image storage part 2, a window image segmenting means 4, a reading means 16 and a collation means 12, the device is provided with a small window image registering means 14 registering a small registered window image with segmenting positional information in a fingerprint dictionary storage part 8 by relating both of them, a reading means reading the small registered window image and the segmenting positional information on the small registered window image from the fingerprint dictionary storage part 8 when the collation error in the collation means 12 is below a preliminarily determined threshold, and a small input window extraction means 18 segmenting the small input window image to be designated by the segmenting positional information of the small window image read from the fingerprint dictionary storage part 8 from the input fingerprint image of the binary image storage part 2.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a fingerprint collation method and apparatus for performing fingerprint collation by using the likelihood of representing a feature point in a feature point window image.

In recent years, with the introduction of computers into a wide range of social systems, it is necessary to maintain the security of those systems. For example, when entering a computer room or using a terminal device, an ID card or password that has been conventionally used as a means for confirming whether or not the person is the person still has a problem in ensuring security. Therefore, the use of a fingerprint collation device is being considered as a powerful alternative to it.
Fingerprint matching is used for fingerprints, "for everyone"
This is because it has two major characteristics, "lifelong immutability."

[0003]

2. Description of the Related Art Conventionally, fingerprint collation uses characteristic points of a fingerprint in order to effectively utilize the above-mentioned characteristics.
The characteristic point is a point (branch point) at which a ridge line formed by a series of convex portions of a fine unevenness formed on a fingertip is divided into two (branch point) or a dead end point (end point). These feature points form the two major features of the fingerprint.

In the fingerprint collation described above, it is possible to increase the speed and flexibly cope with the difference in how the force is applied when the fingerprint is input and the partial expansion and contraction of the fingerprint image caused by moving the glass with the finger attached. What can be done is a problem to be solved. In order to meet this demand, the applicant has already proposed a fingerprint registration / verification method called “moving window method”. The outline of this fingerprint registration / verification method is as follows.

In this fingerprint registration / collation method, one window for alignment and a plurality of feature point window images for collation that are acquired so as to surround the window for alignment are registered in the feature information storage unit. Let me register. FIG. 18 shows an example in which the one alignment feature point window image and a plurality of matching feature point window images are shown on the input fingerprint image. The registration process will be described with reference to FIGS. 19 to 21. The multi-valued fingerprint image signal from the fingerprint sensor 20A is stored in the multi-valued image storage unit 21 and binarized by the multi-valued image binarization unit 22. It is stored in the binary image storage unit 24. Two of the fingerprints in the binary image storage unit 24
The value image is thinned by the image thinning means 28 and written in the thin line image storage unit 30. Then, the feature extraction means 32 extracts feature points. The binarized image of a predetermined range in the vicinity of the extracted predetermined number of characteristic points is cut out as a "window" by the window image cutting means 34 and the window is registered in the characteristic information storage unit 38 (FIG. 2).
2).

At the time of registration, as registration information,
The binary image in "window", the relative positional relationship between "window" and "window", the types of feature points in "window", and the like are registered in the feature information storage unit 38 and are also registered. Select one of the "windows" that is stable and relatively close to the center of the fingerprint and define it as the "registration registration feature point window image" and the rest as "registration matching feature point window image". register.

Further, at the time of the registration, personal information (ID number, etc.) is input from the ten-key pad 20B provided on the fingerprint sensor 20A, and the registration position alignment feature point window image and registration verification feature are input to the personal information recording section 42. Register the window image in association with personal information.

Next, when the "matching" process is started, the multi-valued fingerprint image signal is output from the fingerprint sensor 20A to the multi-valued image storage unit 2.
After being stored in 1 (see S1 in FIG. 23), the fingerprint image is binarized by the multi-valued image binarizing means 22 (see S2 in FIG. 23) and stored in the binary image storage unit 24. . And
Coordinates of the "registered position alignment feature point window image" registered in the feature information storage unit 36 (usually window cutout coordinates are used).
Is used by the alignment window scanning means 46 to store the feature information storage unit 3
The “registered position alignment feature point image” is read from 8.
The "registered position alignment feature point window image" and the binary image storage unit 24 are written as described above, and the memory address designating unit 48 receives the coordinates from the feature information storage unit 38.
The pixel value comparison means 50 performs pattern matching with the window image for registration in the input fingerprint image read by the pixel value comparison means 50 (see S3 in FIG. 23), and the displacement amount indicating the difference between the input positions at the time of registration and at the time of matching. To extract. In this extraction,
If the matching error is not less than the threshold value (S in FIG.
No. 4), until the matching error becomes less than or equal to the threshold value (see Yes in S4 of FIG. 23), the position of the “registered position alignment feature point window image” is moved spirally by 1 or 2 dots. Matching is performed to determine whether the movement is within a predetermined movement range (see S5 in FIG. 23). If it is within the movement range (see Yes in S5 of FIG. 23), the "registration registration feature point window image" is moved, and the registration target window image in the input fingerprint image for that "registration registration feature point window image" is moved. Pattern matching with (Fig. 2
3 S6, S3). When it is determined that the movement is within a predetermined movement range, if the movement is outside the predetermined range (see FIG. 2).
3 (see No in S5), the fingerprint collation for the input image is rejected and the fingerprint collation is rejected (see S7 in FIG. 23).

If the matching error is below the threshold value (see FIG.
3 refer to Yes in S4), the displacement amount is used for matching the registered matching feature point window image with the matching target window image in the input fingerprint image. That is, the pattern matching (comparison between the registered matching feature point window image moved by the displacement amount and the input fingerprint window image corresponding to the registered matching feature point window image, that is, the matching target window image in the input fingerprint image. ) Is performed (S8 in FIG. 25)
reference). In this pattern matching, if the matching error is not less than the threshold value (see No in S9 of FIG. 25), a predetermined number (for example, 1 or 2) of pixels (dots) are spirally formed in the registration matching feature point window image. Only the registration matching feature point window image is moved by the matching window scanning means 44 (see FIG. 24), and if the movement is within a predetermined moving range (see Yes in S10 of FIG. 25), the registration matching feature is displayed. The dot window image is moved by the displacement amount ((S11 of FIG. 25.
Although the pattern matching for the registration matching feature point window image is started (see FIG. 26), if it is outside the movement range (see No in S10 of FIG. 25), another registration matching feature point window image is displayed. Is selected (see S12 of FIG. 25) and the process of step S8 is started. In the matching between the selected registered matching feature point window image and the corresponding matching target window image in the input fingerprint image, the displacement corresponding to the displacement amount is given to the registered matching feature point window image. In the spiral movement processing of the registered matching feature point window image by the matching window scanning means 44, since the human finger is very soft, the input fingerprint image is partially changed due to the difference in the force applied at the time of pressing. This is to absorb the expansion and contraction.

In the matching (pattern matching), the matching is obtained (passed) and the pixel value comparing means 5 is obtained.
The pass window signal output from 0 is counted by the pass window number evaluation means 52, and when the pass number is larger than a predetermined threshold value (see Yes in S13 of FIG. 25), the confirmation of the person is obtained. A signal to perform is output from the pass window number evaluation means 52.

The pass window signal is provided to the system control means 56 via the internal interface 54. The system control means 56 comprises a computer, a work memory, and an external interface. The system control means 56 also receives the name and the like of the fingerprint input person from the personal information storage section 42 via the internal interface 54. The computer of the system control means 56 sends a pass window signal, a name, etc. to the entrance / exit management means 58 as room entry management information via its external interface, and uses them for entry / exit management. Access control 58
Consists of a computer and memory. The computer of the system control means 56 transfers the unlock signal to the door control means (electric lock) via the external interface to unlock the door.

If the number of passes is less than the threshold value (see No in S13 of FIG. 25), and the above-mentioned processing is not completed for all registration matching feature point window images (FIG. 25).
No. of S14), the coordinate information of the next registration matching feature point window image registered in the feature information storage unit 36 is used (see S12 of FIG. 25). The above-mentioned processing, that is, steps S8 to S14
Do.

The above-mentioned processing steps 8 to S14 have been carried out for all the registration matching feature point window images, but if the number of passes still does not reach the threshold value (FIG. 25).
(See Yes in S14), the "registered position alignment feature point window image" is scanned again to find a place where the matching error is equal to or less than the threshold value, and the process of FIG. 25 is performed. If the number of passes still does not reach the threshold value even after the scanning within the defined scanning range is completed, the collation process for the person to be collated is rejected.

[0014]

[Problems to be Solved by the Invention]
The “window method” performs the fingerprint collation processing as described above, and the pattern matching in this method uses the allowable value of the matching error. This allowable value is given in order to deal with a change in ridge width due to a change in finger pressure, rotation of an input fingerprint, and blurring of a pattern.

The allowable value is a value given to the entire feature point window. Therefore, it is difficult to identify the minute difference between the feature point and the normal part. Further, by reducing the allowable value of the matching error, it is possible to identify a fine difference, but in that case, there is a possibility that the collation rate may be lowered.

The present invention has been made in view of the above technical problems, and provides a moving window fingerprint matching method and apparatus capable of identifying the features of the fingerprint pattern in more detail and achieving a reduction in the false matching rate. Its purpose is to provide.

[0017]

1 and 2 show block diagrams of the principle of the invention according to claims 2 and 4, respectively.

The invention according to claim 1 is a fingerprint collation method for collating a registered feature point window image with an input feature point window image, wherein when registering the registered feature point window image, A region with a high probability of representing a feature point is cut out, the region is registered as a registered feature point small window image, and the cut-out position coordinates of the registered feature point small window image are registered in association with the registered feature point small window image, When the matching error in the matching is less than or equal to a predetermined threshold value,
The registered registered small window image and the cutout position coordinates are read out, and the input small window image specified by the cutout position coordinates is cut out from the input feature point window image, and the read registered small window image and the cutout image are cut out. It is characterized in that whether the matching error with the input small window image is less than a predetermined threshold value is used.

According to a second aspect of the present invention, as shown in FIG. 1, the registered feature point window image and the segmented image which are clipped from the binary image storage section 2 by the window image segmentation means 4 and registered in the fingerprint dictionary storage section 8. A fingerprint in which the position coordinates are read in response to the fingerprint image input identification information, and the matching means 12 matches the input feature point window image designated by the cut-out position coordinates in the input fingerprint image with the registered feature point window image. In the collation device, a region with high certainty representing the feature point in the feature point window image cut out by the window image cutout means 4 is cut out, and the region is used as a registered small window image together with the cutout position information of the registered small window image. When the matching error in the small window image registration means 14 for registering both in the fingerprint dictionary storage 8 and the matching means 12 is less than or equal to a predetermined threshold value, the fingerprint dictionary Read-out means 16 for reading the registered small window image and the cut-out position information of the registered small window image from the storage unit 8 and the input specified by the cut-out position information of the registered small window image read out from the fingerprint dictionary storage unit 8. An input small window extracting means 18 for cutting out a small window image from the input fingerprint image of the binary image storage unit 2 is provided, and the registered small window image read by the matching means 12 and the input small window image registration means 18 are provided. Match with the cut out input window image,
It is characterized in that whether or not the matching error between the registered small window image and the input small window image is less than or equal to a predetermined threshold value is used.

According to a third aspect of the present invention, in a fingerprint collation method for collating a registered feature point window image with an input feature point window image, an area with high certainty representing a feature point included in the registered feature point window image. An image is cut out as an image of the registration weighting target area, the image of the registration weighting target area and the cutout position information of the weighting target area are registered in association with each other, and the registered feature point window image and the input feature point window image are collated. When performing, the image of the registration weighting target area,
Regarding collation with the image of the area cut out by the cut-out position information of the weighting target area from the input fingerprint image,
The image of the area other than the registration weighting target area in the registered feature point window image and the image of the area corresponding to the area other than the registration weighting target area in the input fingerprint image are collated and weighted differently. Matching is performed, and whether the matching error between the entire registered feature point window image and the entire input feature point window image is less than or equal to a predetermined threshold value is used.

The invention according to claim 4 is, as shown in FIG. 2, a registered feature point window image and a segmented image extracted from the binary image storage section 2 by the window image segmentation means 4 and registered in the fingerprint dictionary storage section 8. A fingerprint in which the position coordinates are read in response to the fingerprint input person identification information, and the matching means 12 performs matching between the input feature point window image specified by the cut-out position coordinates in the input fingerprint image and the registered feature point window image. In the collation device, an image of a highly probable region representing a feature point included in the registered feature point window image cut out by the window image cutout unit 4 is cut out as an image of a registration weighting target region, and fingerprint image input person identification information correspondence is performed. Registration means 1 for causing the image of the registration weighting target area and the cut-out position information of the registration weighting target area to be registered in the fingerprint dictionary storage unit 8
4A, read-out means 16A for reading the image of the registration weighting target area and cut-out position information of the registration weighting target area in response to the fingerprint image input person identification information from the fingerprint dictionary storage section 8, and the window image cutting-out means. A weighting area designation for designating an area corresponding to the cut-out position coordinate of the weighting target area read by the reading means 16A for the image of the area corresponding to the registered feature point window image cut out from the input fingerprint image by 4. Means 18A
And a registered feature point window image read by the reading means 16A and an image of an area corresponding to the registered feature point window image cut out from the input fingerprint image by the window image cutout means 4. In the collation, for the area designated by the weighting area designating means 18A,
Different weighting is performed by collating the image of the area other than the registration weighting target area in the registration feature point window image with the image of the area corresponding to the image of the area other than the registration weighting target area in the input fingerprint image. However, it is characterized in that whether or not the matching error between the entire registered characteristic point window image and the entire input characteristic point window image is less than or equal to a predetermined threshold value is used.

According to a fifth aspect of the present invention, in the fingerprint collation method according to the first or third aspect, the certainty of representing a feature point is a branched portion of a branch point. .

According to a sixth aspect of the present invention, as shown in FIG. 1, in the fingerprint collation apparatus according to the second or fourth aspect, the likelihood of representing a feature point is a branch point branching portion. Is characterized in that.

According to a seventh aspect of the invention, in the fingerprint collation method according to the first or third aspect, the certainty of representing the characteristic point is a portion prior to the tip portion of the end point. To do.

The invention according to claim 8 is, as shown in FIG. 2, in the fingerprint collating apparatus according to claim 2 or claim 4, the certainty of representing a characteristic point is a portion before the tip portion of the end point. It is characterized by being.

The invention according to claim 9 is characterized in that, in the fingerprint collation method according to claim 1 or 5, collation errors of the registered feature point window image and the registered feature point small window image are individually set. And

According to a tenth aspect of the present invention, as shown in FIG. 1, in the fingerprint collation device according to the second or sixth aspect, the collation error between the registered feature point window image and the registered feature point small window image is eliminated. It is characterized by being set individually.

In the invention according to claim 11, in the fingerprint matching method according to claim 1, claim 5, or claim 9, a small window image is set only for the registration-position feature point window image. Is characterized by.

According to a twelfth aspect of the invention, as shown in FIG. 1, in the fingerprint collating apparatus according to the second, sixth or tenth aspect, only for the registration position feature point window image. It is characterized by setting a small window image.

The invention according to claim 13 is the fingerprint matching method according to claim 1, claim 5, or claim 9, wherein a small window image is set only for at least one registration matching feature point window image. It is characterized by doing.

According to a fourteenth aspect of the present invention, as shown in FIG. 1, in the fingerprint collation device according to the second, sixth, or tenth aspect, at least one feature point window image for registration collation is displayed. The feature is that only a small window image is set.

According to a fifteenth aspect of the present invention, in the fingerprint collation method according to the third or seventh aspect, weighting is set only for the registered position alignment feature point window image.

According to a sixteenth aspect of the present invention, as shown in FIG. 2, in the fingerprint collation apparatus according to the third or seventh aspect, weighting is set only for the registered collation feature point window image. Is characterized by.

[0034]

According to the first aspect of the present invention, when registering a registered feature point window image as in the prior art, an area having a high probability of representing a feature point in the registered feature point window image is set as a registered small window image. Together with the cut-out position coordinates, both are associated and registered so as to be readable. In the invention according to claim 2, this registration is performed by the small window image registration means 14.

The matching between the registration position feature point window image and the registration target window image corresponding to the registered position feature point window image in the input fingerprint image is performed in the same manner as the conventional method. Calculate the amount of displacement.

The matching error between the registered feature point window image moved by this displacement amount and the matching target window image corresponding to the registered feature point window image in the input fingerprint image is a predetermined threshold value. In the following cases, the registered registration window image and the cut-out position coordinates are read. In the invention according to claim 2, this reading is performed by the reading means 16.

Then, the input window image specified by the cut-out position coordinates is cut out from the input fingerprint image. In the invention according to claim 2, this clipping is performed by the input small window extracting means 18.

By collating the read registered small window image with the cut out input small window image, the collation result when the collation error is equal to or less than a predetermined threshold value is confirmed by the person himself / herself. To use.

The invention according to claim 3 uses a weighting target area instead of using the window image of the invention according to claim 1. In the matching for the weighting target area, the matching error is calculated with a weight different from that of the other areas. The other points are almost the same as in claim 1. Therefore, also in claim 4, the constituent elements are referred to by adding the letter A, and the description thereof will not be repeated.

It is to be noted that claims 5, 6, and 9 to 14 limit the constituent elements of claims 1 and 2. Further, claims 7 and 8, and claims 15 and 16 limit the constituents of claims 3 and 4.

[0041]

FIG. 3 to FIG. 5 show an embodiment of the invention according to claim 1, claim 2, claim 5, claim 6, and claim 9 to claim 14. This embodiment is also a fingerprint collation device by the Mubinck window method. In this figure,
The same reference numerals as those in FIGS. 19 to 21 indicate the same components, and thus the details thereof will be omitted. In the embodiment shown in FIG. 3 to FIG. 5, claim 1, claim 2, claim 5,
The features of the invention according to claim 6 and claims 9 to 14 are the ridge direction extracting means 35 in the window image shown in FIG.
It is the feature point orientation extraction means 37 in the window image, the feature point small window image cutout means 39, the feature information storage section 41, and the small window image scanning means 47 shown in FIG.

The in-window image ridge direction extraction means 35 extracts the direction of the ridge or valley line in the window image cut out by the window image cutout means 34. The window image cutting means 34 is shown in FIG.
As shown in (1), the binary fingerprint image is extracted in a window shape with the feature points extracted by the feature point extracting means 32 as the center. The in-window image feature point orientation extraction means 37 extracts the orientation of the feature points in the window image clipped by the window image clipping means 34. The feature point small window image cutting-out means 39 is the type of feature point extracted by the feature point extracting means 32, the direction of the ridge or valley extracted by the window image ridge direction extracting means 35, and the feature point in the window image. An area with a high probability of representing a feature point from the feature point window image by using the orientation of the feature point extracted by the orientation extraction means 37 [a portion in which the feature point in the feature point window image is clearly represented (the tip of the end point is The crotch part of the branch point)] is cut out as a registration matching feature point small window image as shown in FIG. The cut out registration matching feature point small window image is cut out and the feature information storage unit 4 together with the cut-out position coordinates of the registration matching feature point small window image.
Registered in 1. In the feature information storage unit 41, the registered position matching feature point window image and the registered matching feature point window image, and their cut-out position coordinates are also registered, as in the conventional case. Window image ridge direction extraction means 35, window image feature point orientation extraction means 37, and feature point small window image cutout means 3
9 is an image thinning unit 28, a thin line image storage unit 30,
The registration unit 27 is configured with the feature extracting unit 32 and the window image extracting unit 34. Further, the characteristic information storage unit 41 constitutes the fingerprint dictionary recording unit 40 together with the personal information storage unit 37.

The small window image scanning means 47 shown in FIG. 4 gives the cut-out position coordinates of the registered matching feature point small window image registered in the feature information storage section 41 to the memory address designating means 48, and the binary image at the time of matching. A matching target small window image (input small window image) corresponding to the registered registered matching feature point small window image in the input binary fingerprint image stored in the storage unit 24 is cut out and the matching target small window image is extracted. A process of inputting the pixel value comparing means 50 together with the registered feature point window image is performed. The small window image scanning means 47 includes a collation window scanning means 44,
Alignment window scanning means 46, memory address designating means 48, pixel value comparing means 50, and passing window number evaluating means 52.
Together with this, the collation unit 43 is configured.

In FIGS. 3 to 5, the binary image storage unit 2
Reference numeral 4 corresponds to the binary image storage unit 2 in FIG. 1, and window image cutout unit 34 corresponds to the window image cutout unit 4 in FIG. The feature information storage unit 41 corresponds to the fingerprint dictionary storage unit 8 in FIG. 1, and the pixel value comparison unit 50 corresponds to the matching unit 12 in FIG. Window image ridge direction extraction means 35, window image feature point orientation extraction means 37, feature point small window image cutout means 3
Reference numeral 9 corresponds to the small window image registration means 14 in FIG. 1, and small window image scanning means 47 corresponds to the reading means 16 in FIG.
Small window image scanning means 47, memory address designating means 48
Corresponds to the input window extracting means 18 in FIG.

The operation of the apparatus having such a configuration will be described below with reference to FIGS. In the registration of the window information, the multi-valued fingerprint image signal read by the fingerprint sensor 20A is stored in the multi-valued image storage unit 21 (see S1 of FIG. 6), as described in the [Prior Art] section. ), It is binarized by the multi-valued image binarization means 22 (see S2 in FIG. 6) and stored in the binary image storage unit 24. The input image is thinned by the image thinning unit 28 (see S3 in FIG. 6) and written in the thin line image storage unit 30 and used by the feature extracting unit 32 to extract the feature points (S4 in FIG. 6). reference). Using the feature points, the window image cutout unit 34 extracts a feature point window image from the binary fingerprint image centering on the feature points extracted by the feature point extraction unit 32 as shown in FIG. One of the window images is registered as the registration position feature point window image, and the other is registered as the registration matching feature point window image in the feature information storage unit 41 (see S5 in FIGS. 9 and 6).

The in-window image ridge direction extraction means 35 uses the registered matching feature point window image cut out by the window image cutting means 34 to determine the direction of the ridge or valley line in the registered matching feature point window image. Is extracted (see S6 in FIG. 6). The in-window image feature point orientation extraction means 37 extracts the orientation of the feature points in the registered collation feature point window image cut out by the window image cutout means 34 (see S7 in FIG. 6). The feature point small window image cutout unit 39 is a type of feature point extracted by the feature point extraction unit 32.
From the registered collation feature point window image using the direction of the ridge or valley line extracted by the window image ridge direction extraction means 35 and the orientation of the feature point extracted by the window image feature point orientation extraction means 37 FIG. 10 shows a region having a high probability of representing a feature point [a part clearly showing the feature in the registration matching feature point window image (the tip of the end point, the tip of the branch point, the crotch part of the branch point)]. As described above, it is cut out as a registration matching feature point small window image. In this way, the cut-out registration-comparison feature point small-window image is the ID input from the ten-key 20B in the feature information storage unit 41 together with the cut-out position coordinates of the registration-comparison feature-point small window image.
Register in correspondence with the number (fingerprint image input person identification information). or,
The registration matching feature point window image and the registration matching feature point window image cut out as described above are registered in the feature information storage unit 41 together with the respective cut-out position coordinates (see S9 in FIG. 6). .

Fingerprint dictionary recording section 4 in fingerprint collation processing
Reading the registration position feature point window image from 0,
The reading of the registration target window image corresponding to the registered registration feature point window image from the binary image storage unit 24 is performed under a known moving window method. That is,
The registration position feature point window image corresponding to the ID number (personal information) input from the numeric keypad 20B is read from the fingerprint dictionary recording unit 40. Along with the reading of the registered position alignment feature point window image corresponding to the ID number, the cut-out position coordinates read from the feature information storage unit 41 are passed to the alignment scan unit 46, and sequentially from the memory address designation unit 48. The binary image storage unit 24 is read at the output memory address. It corresponds to the registration position feature point window image read from the fingerprint dictionary recording unit 40 and the registration position feature point window image in the input fingerprint image read from the binary image storage unit 24. The pattern matching with the target window image for alignment is performed by the pixel value comparison means 5.
At 0.

The pattern matching is carried out in steps S1 to S6 of FIG. 7. Since these steps are the same as steps S1 to S6 of FIG. 23, the steps described in steps S1 to S6 of FIG. Done. By this registration processing, the amount of displacement between the registered registration feature point window image and the registration target window image corresponding to the registered registration feature point window image in the input fingerprint image is obtained.

In the pattern matching (matching) between the registration position matching feature point window image and the position matching target window image in the input fingerprint image, if the position matching is possible, the registration matching feature point window is the same as the conventional one. The image is compared with the matching target window image corresponding to the registered matching feature point window image in the input fingerprint image.

The collation is exactly the same as that described with reference to FIG. 25 except for the differences described below. The difference is the processing of steps S15 and S16 shown in FIG. Other steps S8, S9, S10, S11, S
25, and the processing is similar to that described with reference to FIG. 25. The characteristic information storage unit 41, the collation window scanning means 44, the memory address designating means 48, 2 This is performed by the value image storage unit 24 and the pixel value comparison means 50. Therefore, the above process will not be repeated, but the above difference will be described below.

The matching error in the pattern matching between the registered matching feature point window image and the matching target window image corresponding to the registered matching feature point window image in the input fingerprint image has a predetermined threshold value. If it is below (see YES in S9 of FIG. 8), the registration matching feature point small window image,
Pattern matching with the small window image to be matched is performed.
The matching target small window image corresponds to the registered feature point small window image in the registration target window image which is cut out from the input fingerprint image and subjected to the pattern matching as described above. This pattern matching process is performed by the feature information storage unit 41, the small window image window scanning unit 47, the memory address designating unit 48, the binary image storage unit 24, and the pixel value comparison unit 50.
However, if the matching error in the pattern matching is not less than or equal to the predetermined threshold value (see NO in S16 of FIG. 8), the registration matching feature point window image and the matching target window in the input fingerprint image are compared. Similar to the pattern matching with the image, if the previous spiral movement of the registration matching feature point window image by a predetermined pixel unit is within the movement range (see YES in S10 of FIG. 8), The movement is caused, and the pattern matching is performed again for the registered registration feature point window image after the movement. If the movement is out of the movement range (see NO in S10 of FIG. 8), similar to the pattern matching between the registration matching feature point window image and the matching target window image in the input fingerprint image, another registration matching is performed. The selection of the feature point window image for use is started (see S12 in FIG. 8).

In the pattern matching in which the registered collation feature point window image is moved in a spiral manner as described above, if the matching error (collation error) falls below a predetermined threshold value (see FIG. 8), refer to YES in S16), the pass window signal is the pass window number evaluation means 52 as in the conventional case.
And add 1 to the count value of the acceptance window,
If the pass window number exceeds a predetermined threshold value (see YES in S13 of FIG. 8), a signal that the person's confirmation is obtained is output from the pass window number evaluating means 52.

The personal identification signal is notified from the pass window number evaluation means 52 to the system control means 56 via the internal interface 54 in the same manner as in the conventional case, and is used for the same entry / exit management as in the conventional case.

In the above-mentioned processing, if the person cannot be confirmed (see No in step S13 in FIG. 8) and all of the registered matching feature point window images are not used in the matching (No in step S14 in FIG. 8). Then, the process proceeds to step S12, and the same process as described above is performed for the next registration matching feature point window image. Thereby, in any one of the registration matching feature point window image matching processing (pattern matching processing),
If the person can be confirmed (see Yes in step S13 of FIG. 8), the person currently being collated is confirmed as the person. On the contrary, if all of the registered matching feature point window images are used for matching (see Yes in step S14 in FIG. 8), the registered registration window image is scanned again within the set range,
If the matching error is less than or equal to the threshold value, the above-mentioned processing is repeated, and when the scanning of the entire range in which the registered registration window image is set is completed, the matching person is rejected because the person cannot be confirmed. Next, claim 3, claim 4, claim 7, claim 8, claim 15, and claim 16
One embodiment of the invention according to the present invention will be described below. In this embodiment, the inventions according to claim 1, claim 2, claim 5, claim 6, and claim 9 to claim 14 have the certainty of representing the feature points in the registration matching feature point window image. Highly-registered matching feature points Instead of using a small window image for matching, a highly probable region representing a feature point in a registered matching feature point window image [a part that clearly represents a feature (a tip of an end point (A tip of a part, a crotch portion of a branch point)] (hereinafter, referred to as a weighting target area) is configured to be weighted. For the weighting, if a normal count value of 1 to 1 is given between pixels in regions other than the weighting target region for determining matching error (matching error), each pixel in the weighting target region is determined. For, for example, the weighting is performed by giving a count value of 1 to 2. The weighting is uniform,
It may be changed if necessary. Only the difference is claim 3,
It is different from one embodiment of the invention according to claim 4, claim 7, claim 8, claim 15 and claim 16. The difference will be described below.

The construction of an embodiment of the invention according to claim 3, claim 4, claim 7, claim 8, claim 15 and claim 16 is shown in FIG. 11 to FIG. Two
Since 3 is the same as the registration unit 23 in FIG. 3, its reference number is also given the same reference number. The registration unit 27A has the same configuration as that of the registration unit 27 shown in FIG. 3, except that the window image feature point orientation extraction unit 37 is connected to the window image weighted region setting unit 39A. As such, each of the components is given the same reference numeral. Since the fingerprint dictionary recording unit 40A is the same except for the characteristic information storage unit 41A, the reference numbers are also attached to the reference numbers. Since the collation unit 43A has the same configuration as each of the components of the collation unit 43 shown in FIG. 3 except for the weighting area designating means 47A, the same reference numerals are assigned to the respective components. I am doing it.

11 to 13, the binary image storage unit 24 corresponds to the binary image storage unit 2 of FIG. 2, and the window image cutout unit 34 corresponds to the window image cutout unit 4 of FIG. The feature information storage unit 41A is the registration means 14A of FIG.
The pixel value comparison means 50 corresponds to
Corresponding to, the in-window image ridge direction extraction means 35, the in-window image feature point direction extraction means 37, and the in-window image weighted area setting means 39A correspond to the registration means 14A in FIG. The weighting area designating means 47A corresponds to the reading means 16A shown in FIG. 1, and the weighting area designating means 47A memory address designating means 48 corresponds to the weighting area designating means 18A shown in FIG.

The in-window image weighting area designating means 39A is
Types of feature points (end points and branch points) extracted by the feature point extraction means 32, directions of ridges or valleys extracted by the window image ridge direction extraction means 35, and window image feature point orientation extraction means 37. An area with a high probability of representing a feature point from the registration matching feature point window image by using the orientation of the feature point extracted in (a part that clearly represents the feature in the registration matching feature point window image (the tip of the end point Image of the weighting target area (image of weighting target area)
Say. ] Is cut out as shown in FIG. Then, the weighting target area image is registered as a registration weighting target area image in the feature information storage unit 41A together with the cut-out position coordinates of the weighting target area image. In the feature information storage unit 41A, as in the embodiments of the inventions according to claim 1, claim 2, claim 5, claim 6, and claim 7 to claim 14, one registration registration feature point. A window image and a plurality of registration matching feature point window images are registered together with their cut-out position coordinates.

A registration position matching feature point window image for performing such fingerprint matching after registration and a registration target window region image corresponding to the registration position matching feature point window image in the input fingerprint image. The collation is the same as that described with reference to FIG. 7 as shown in steps S1 to S6 of FIG. 15, and thus the detailed description thereof will be omitted. As described above, this collation is performed by the ten-key pad 20A, the fingerprint dictionary recording unit 40A, the alignment window scanning unit 46, the memory address designating unit 48, the binary image storage unit 24, and the pixel value comparing unit 50. By this alignment, the amount of displacement between the registered registration feature point window image and the registration target window image corresponding to the registered registration feature point window image in the input fingerprint image is obtained.

When performing matching between the registered matching feature point window image using the displacement amount performed after this alignment and the matching target window image in the input fingerprint image corresponding thereto, The matching between the registration weighting target area image and the matching target window area image corresponding to the registration weighting target area image in the input fingerprint image is the difference described below, that is, FIG.
8 is the same as that described with reference to FIG. 8 except for the processing of steps S15 and S16.

When the matching is started, the image of the area other than the registration weighting target area in the registration matching feature point window image and the alignment target corresponding to the area other than the registration weighting target area in the input fingerprint image are registered. In the matching with the window image, pattern matching is performed for each pixel without weighting as in the conventional method, but in the matching between the registration weighting target area image and the area image corresponding to the registration weighting target area image in the input fingerprint image, Collation is performed by giving the weighting to the pixels as described above (see S15 in FIG. 16). If the matching error is not less than or equal to the predetermined matching error (see YES in S16 of FIG. 16), the spiral previous movement of the registration matching feature point window image by a predetermined pixel unit is within the movement range. If so (see YES in S10 of FIG. 16), the pattern matching is performed again by causing the movement and giving the above-mentioned weighting to the registration matching feature point window image after the movement. If the movement is out of the movement range (see NO in S10 of FIG. 16), another registration matching feature point window image is selected as in the conventional case (FIG. 16).
See S12).

As described above, in the pattern matching in which the registration matching feature point window image is moved spirally,
If the matching error becomes equal to or smaller than a predetermined threshold value (see YES in S16 of FIG. 16), the pass window signal is given to the pass window number evaluation means 52 as in the conventional case, and the pass window count value is set to 1 Just add. If the number of passing windows exceeds a predetermined threshold value (YES in S13 of FIG. 16)
(See reference), a principal confirmation signal that the confirmation of the principal is obtained is output from the pass window number evaluation means 52.

The principal confirmation signal is notified from the pass window number evaluation means 52 to the system control means 56 via the internal interface 54 in the same manner as in the conventional case, and is used for the same entry / exit management as in the conventional case.

In the above-mentioned processing, if the person cannot be confirmed (see No in step S13 of FIG. 16) and all of the registered matching feature point window images are not used for matching (see FIG. 1).
6 (see No in step S14), the process proceeds to step S12, and the same process as described above is performed for the next registration matching feature point window image. As a result, if the person can be confirmed in the matching process (pattern matching process) of any of the registration matching feature point window images (Yes in step S13 of FIG. 16).
(See s), and confirm the person who is currently performing the verification process as the person.
Conversely, if all of the registered matching feature point window images are used for matching (see Yes in step S14 of FIG. 16), the matching process is rejected because the person cannot be confirmed. Then, the alignment window is scanned again to find a new alignment location. If a new alignment location is found, the above process is repeated. When the scanning of the predetermined scanning range is completed, the process ends.

In the two embodiments, the matching error is less than or equal to a predetermined threshold value for the registered feature point window images in each registration matching feature point window image, or the weighting target area is set. The example in which the matching error of the registration matching feature point window image weighted with is less than or equal to a predetermined threshold value has been described, but the same applies to all feature point window images or alignment feature point window images. May be applied.

Further, the registration matching feature point window image and the corresponding matching target window image in the input fingerprint image are matched, and the registration matching feature point small window image and the corresponding input fingerprint image are also matched. The matching errors in the matching with the matching target small window image are the same, but different matching errors may be used. The registration matching feature point window image may be provided only for the alignment feature point window image, or may be provided for at least one registration matching feature point window image.

The weighting target area may be provided only for the alignment feature point window image, or may be provided for at least one registration matching feature point window image.

[0067]

As described above, according to the present invention, at the time of fingerprint collation, the collation is performed with respect to a highly probable region representing a feature point in a feature point window image.
Since the features of the fingerprint pattern can be identified in more detail, the false matching rate can be reduced. When the false matching rate is set to the same level as the conventional one, the allowable value of the matching error can be increased, so that the fingerprint pattern is less likely to change.

[Brief description of drawings]

1] Claims 2, 6, 6, 10
FIG. 15 is a principle block diagram of an invention according to claim 14;

FIG. 2 Claims 4, 6, 8, and 1
It is a principle block diagram of the invention which concerns on 6.

FIG. 3 is a diagram showing a first part obtained by dividing an embodiment of the invention according to claim 1, claim 2, claim 5, claim 6, and claim 9 to claim 14 into three parts.

FIG. 4 is a diagram showing a second part obtained by dividing an embodiment of the invention according to claim 1, claim 2, claim 5, claim 6, and claim 9 to claim 14 into three parts.

FIG. 5 is a diagram showing a third part obtained by dividing the embodiment of the invention according to claim 1, claim 2, claim 5, claim 6, and claim 9 to claim 14 into three parts.

FIG. 6 is a diagram showing a registration processing flow of an embodiment of the invention according to claim 1, claim 2, claim 5, claim 6, and claim 9 to claim 14;

FIG. 7 is a diagram showing a part of a collation processing flow of an embodiment of the invention according to claim 1, claim 2, claim 5, claim 6, and claim 9 to claim 14;

FIG. 8 is a diagram showing the rest of the collation processing flow of an embodiment of the invention according to claim 1, claim 2, claim 5, claim 6, and claim 9 to claim 14;

FIG. 9 is a diagram showing the registered registration feature point window image to registration feature point window image.

FIG. 10 is a diagram showing an extracted registration matching feature point window image and registration matching feature point window image.

FIG. 11: Claim 3, Claim 4, Claim 7, Claim 8,
It is a figure which shows the 1st part which divided one Example of invention of Claim 15 and Claim 16 into 3 parts.

FIG. 12: Claim 3, Claim 4, Claim 7, Claim 8,
It is a figure which shows the 2nd part which divided one Example of invention of Claim 15 and Claim 16 into 3 parts.

FIG. 13 is a combination of claim 3, claim 4, claim 7, claim 8,
It is a figure which shows the 3rd part which divided one Example of invention of Claim 15 and Claim 16 into 3 parts.

FIG. 14 is a combination of claims 3, 4, 7 and 8;
It is a figure which shows the registration processing flow of one Example of the invention which concerns on Claim 15 and Claim 16.

FIG. 15 is a combination of claims 3, 4, 7 and 8;
It is a figure which shows a part of collation process flow of one Example of the invention which concerns on Claim 15 and Claim 16. FIG.

FIG. 16 is a combination of claim 3, claim 4, claim 7, claim 8, and
It is a figure which shows the remainder of the collation processing flow of one Example of the invention which concerns on Claim 15 and Claim 16.

FIG. 17 is a diagram showing weighting target areas set in a registration matching feature point window image.

FIG. 18 is a diagram showing an input fingerprint image, an alignment feature point window image, and a matching feature point window image.

FIG. 19 is a diagram showing a first part obtained by dividing a conventional moving window fingerprint matching device into three parts.

FIG. 20 is a diagram showing a second part obtained by dividing a conventional moving window fingerprint matching device into three parts.

FIG. 21 is a diagram showing a third part obtained by dividing a conventional moving window fingerprint matching device into three parts.

FIG. 22 is a diagram showing an example of registration in a characteristic information storage unit.

FIG. 23 is a diagram showing a part of a matching process flow of a conventional moving window fingerprint matching device.

FIG. 24 is a diagram showing how registration is performed during a matching process on one registered matching feature point window image.

FIG. 25 is a diagram showing the rest of the matching process flow of a conventional moving window fingerprint matching device.

FIG. 26 is a diagram showing how registration is performed in a matching process when a registered matching feature point window image is selected.

[Explanation of symbols]

 2 Binary image storage unit 4 Window image cutout unit 8 Fingerprint dictionary storage unit 10 Readout unit 12 Collation unit 14 Small window image registration unit 14A Registration unit 16 Readout unit 16A Readout unit 18 Input small window extraction unit 18A Weighted area setting unit 24 2 Value image storage unit 34 Window image cutout unit 35 Window image ridge direction extraction unit 37 Window image feature point direction extraction unit 39 Feature point small window image cutout unit 39A Window image weighted area setting unit 41 Feature information storage unit 41A Features Information storage section 44 Collation window scanning means 47 Small window image scanning means 47A Weighting area designating means 48 Memory address designating means 50 Pixel value comparing means

Claims (16)

[Claims] 1. A fingerprint collation method for collating a registered feature point window image with an input feature point window image, wherein at the time of registering the registered feature point window image, the probability of representing a feature point in the registered feature point window image. Of the high feature area is registered and the area is registered as a registered feature point small window image, and the cut-out position coordinates of the registered feature point small window image are registered in association with the registered feature point small window image. When it is less than or equal to a predetermined threshold value, the registered registered feature point window image and cut-out position coordinates are read, and the input window image specified by the cut-out position coordinate is cut out from the input feature point window image. A finger characterized by using whether the matching error between the read registration matching feature point small window image and the cut out input small window image is less than or equal to a predetermined threshold value. Crest matching method. 2. A fingerprint image storage unit (8) which is cut out from a binary image storage unit (2) by a window image cutout unit (4).
The registered feature point window image and the cut-out position coordinates registered in the above are read out in response to the fingerprint image input identification information, and the input feature point window image designated by the cut-out position coordinates in the input fingerprint image and the registered feature point window are read. In a fingerprint collation device for collating with an image, a region with high certainty representing a feature point in a registered feature point window image cut out by the window image cutting means (4) is cut out and the region is registered as a feature point small window. There is a matching error between the small window image registering means (14) for registering in the fingerprint dictionary storage section (8) in association with the cut-out position information of the registered feature point small window image as an image, and the matching error in the matching means (12). When it is less than or equal to a predetermined threshold value, a reading means (1) for reading the registered feature point small window image and the cut-out position information of the registered feature point small window image from the fingerprint dictionary storage unit (8). ) And an input small window image designated by the cut-out position information of the registered feature point small window image read from the fingerprint dictionary storage unit (8) is cut out from the input fingerprint image of the binary image storage unit (2). An input small window extracting means (18) is provided to match the registered small window image read by the matching means (12) with the input small window image cut out by the input small window image registration means (18). The fingerprint collation apparatus is characterized in that the collation error between the registered small window image and the input small window image is less than or equal to a predetermined threshold value. 3. A fingerprint collation method for collating a registered feature point window image with an input feature point window image, wherein an image of a highly probable region representing a feature point included in the registered feature point window image is a registration weighting target region. The image of the registration weighting target area, which is cut out as an image, and the cutout position information of the registration weighting target area are registered in association with each other, and when the registered feature point window image and the input feature point window image are collated, Regarding the matching of the image of the registration weighting target area and the image of the area cut out by the cut-out position information of the weighting target area from the input fingerprint image, an area other than the registration weighting target area in the registered feature point window image And the image of the area corresponding to the area other than the registration weighting target area in the input fingerprint image are weighted differently from each other to perform the matching. Moving window fingerprint verification method characterized by using either matching error between the registered feature point windows entire image and the input feature point window entire image is a predetermined threshold or less. 4. A fingerprint image storage unit (8) which is cut out from the binary image storage unit (2) by a window image cutout unit (4).
The registered feature point window image and the cut-out position coordinates registered in the above are read in response to the fingerprint input person identification information, and the input feature point window image designated by the cut-out position coordinates in the input fingerprint image and the registered feature point window are read. Collating means (12) for collating with an image
In the fingerprint collating apparatus according to the present invention, an image of an area with high probability that represents a feature point included in the registered feature point window image cut out by the window image cutout means (4) is cut out as an image of the registration weighting target area, and the fingerprint image Registration means (14A) for registering the image of the registration weighting target area and the cutout position information of the registration weighting target area in the fingerprint dictionary storage section (8) corresponding to the input person identification information, and the fingerprint dictionary storage section (8). A read means (16A) for reading the image of the registration weighting target area and the cutout position information of the registration weighting target area in response to the fingerprint image input person identification information from the input fingerprint image by the window image cutting means (4). The image of the region corresponding to the registered feature point window image cut out from is read out by the reading means (16A). Weighting area designating means (18A) for designating an area corresponding to the clipping position coordinate of the weighting target area is provided, and the registered feature point window image read by the reading means (16A) and the window image cutting means. In the matching with the image of the area corresponding to the registered feature point window image cut out from the input fingerprint image by (4), the area specified by the weighting area setting means (18A) is the registered feature point window. The image of the region other than the registration weighting target region in the image and the image of the region corresponding to the image of the region other than the registration weighting target region in the input fingerprint image are weighted differently, and the registration feature A fingerprint collation device characterized by using whether or not a collation error between the entire point window image and the entire input feature point window image is equal to or less than a predetermined threshold value. 5. The fingerprint matching method according to claim 1 or 3, wherein the likelihood of representing a feature point is a branched portion of a branch point. 6. The fingerprint collation apparatus according to claim 2, wherein the likelihood of representing a feature point is a branched portion of a branch point. 7. The fingerprint matching method according to claim 1 or 3, wherein the likelihood of representing a feature point is a portion prior to the tip portion of the end point. 8. The fingerprint collation apparatus according to claim 2 or 4, wherein the likelihood of representing a feature point is a portion of the endpoint that is ahead of the tip portion. 9. The fingerprint matching method according to claim 1, wherein a matching error between the registered feature point window image and the registered feature point small window image is set individually. 10. The fingerprint collation apparatus according to claim 2, wherein the collation error between the registered feature point window image and the registered feature point window image is set individually. 11. The fingerprint collation method according to claim 1, 5, or 9, wherein a small window image is set only for a registration position feature point window image. Method. 12. The method according to claim 2, claim 6, or claim 10.
The fingerprint collation device according to the item (1), characterized in that a small window image is set only for a registration position feature point window image. 13. The fingerprint matching method according to claim 1, claim 5, or claim 9, wherein a small window image is set only for at least one registration matching feature point window image. Fingerprint matching method. 14. The method according to claim 2, claim 6, or claim 10.
2. The fingerprint collation device according to the item (1), wherein a small window image is set only for at least one registration-use feature point window image. 15. The fingerprint collation method according to claim 3 or 7, wherein weighting is set only for the registration position feature point window image. 16. The fingerprint collation apparatus according to claim 3, wherein weighting is set only for the registration collation feature point window image.