JPH04324583A - Method and device for moving-window type fingerprint picture collation - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a moving window fingerprint image matching method and device, and more particularly to a moving window type fingerprint image matching method and matching device, and more particularly to a method that enables identification even when one alignment window does not match.

0002

BACKGROUND OF THE INVENTION Passcard or password systems exist for identifying individuals who are authorized to enter and exit critical facilities, such as computer rooms or bank terminal rooms. The former is problematic when you forget your card, and the latter is problematic when you can't remember your password. For this purpose, one of the means of personal identification is
One example is a fingerprint verification system.

[0003] In this fingerprint verification system, an individual's fingerprint is handled in advance as an image, as shown in FIG. 7A, for example, and features are extracted and verified based on this image. Therefore, it consists of an input means for converting a fingerprint into image data, and a collation means for comparing the registered image data with the input fingerprint image data.

[0004] When registering, as shown in FIG. 7(A),
As shown in FIG. 7(B), the alignment window W0 and the matching windows W1 to W6 are extracted from the fingerprint ridge image, and each window W0 is
~W6 is registered in the fingerprint dictionary.

[0005] Alignment window W0 and verification windows W1 to W6
uses fingerprint minutiae, such as branch points and endpoints.

[0006] These windows are first extracted, for example, from the alignment window W0. The alignment window is extracted as follows. That is, as shown in FIG. 7(A), when a feature point is first extracted by scanning from the center of the fingerprint ridge image in a spiral shape (either right or left is fine), it is used as the alignment window W0. do.

[0007] When extracting feature points, a multivalued input image is binarized and thinned to a 1-bit width, and this is scanned using a 3×3 window, for example, to detect end points and branch points. After detecting the feature point in this way, centering on this point,
A window with a size of, for example, 16×16 pixels or 32×32 pixels is applied to the fingerprint ridge image to obtain a wide binary image alignment window W0 as shown in FIG. 6(B).

The matching windows W1 to W6 are extracted in the same manner. That is, after extracting the position of the alignment window W0, spiral scanning is further continued using the 3×3 window. When a minutiae point is detected, the above-mentioned large-sized window is applied to the fingerprint ridge image as shown in FIG. 7(A) to obtain a matching window W1 centered on the minutiae point.

[0009] In this way, W1, W2, . . . are sequentially extracted, and a predetermined number of matching windows are extracted. Of course, Figure 7 (B
), the alignment window W0 and the verification windows W1 to W6
Each image data is registered as personal information in the fingerprint dictionary together with position data.

[0010] Fingerprint verification is performed by pattern matching of fingerprint image data of a large amount of personal information stored in this manner. At this time, the inside of the fingerprint dictionary is divided into one alignment window W0 and a plurality of matching windows W1 to W6 around the alignment window W0.

The specific order of verification will be explained according to the flowcharts for explaining the verification operation shown in FIGS. 7(C) and 8.

[0012] First, pattern matching is performed on the entire fingerprint input during verification with the alignment window W0. At this time, matching is performed based on the position data of the alignment window W0, but if the matching degree of the pattern does not reach the standard value and it cannot be determined that it is a candidate part of the alignment window, the alignment window W0 is moved in a spiral shape. and perform pattern matching. At this time, pattern matching is performed for objects whose positional movement amount is within a predetermined movement range, but if a match is still not obtained and the movement exceeds the specified movement range, input The fingerprints received will be judged to be different from those of the registrant and will be rejected.

■The amount of movement of the alignment window is within the specified range,
When a match is obtained for the pattern of the alignment window W0, each of the matching windows W1 to W6 is moved by the amount of movement.

[0014] Then, matching is performed for the patterns of each of the matching windows W1 to W6. At this time, for example, the fingerprints are compared sequentially starting from the one closest to the center of the fingerprint. In this example W5
, W2, W6, . . . individually in this order. In this case as well, the alignment window is moved two-dimensionally in a spiral manner by a predetermined amount, centering on the position where the alignment window has been moved by the amount of movement described above, in order to accommodate distortions caused by the softness of human skin, for example. and perform pattern matching.

[0015] In this way, when first matching W5, only 1 out of 6 passed, so if the number of passes is, for example, 4, it is determined that it has been confirmed, this confirmation judgment The threshold has not been reached. Therefore, the next W
Verification processing is performed for 2. In this way, when the verification constant for passing reaches the threshold value, it is determined that the identity has been verified. However, when the verification process is completed for the entire verification window without reaching the threshold, the input fingerprint is determined to be different from that of the registrant and is rejected.

FIG. 7C shows a matching state between the matching image and the dictionary image, that is, the pattern of the alignment window and the pattern of the matching window. Note that in FIG. 7(C), W0 to W6 are expressed only by numbers to simplify the illustration.

[0017]

[Problems to be Solved by the Invention] In the conventional case, as shown in FIGS. 7 and 8, pattern matching is first performed for the alignment window W0, and then each matching window W1, W
2..., so if the pattern matching of the alignment window W0 was not established, subsequent matching could not be performed.

In other words, when pattern matching is performed on the alignment window, if matching is not possible for some reason, for example, if the alignment location cannot be found due to blurred fingerprint ridges, poor reproducibility, etc., matching is not possible. It may not be possible.

SUMMARY OF THE INVENTION Therefore, an object of the present invention is to provide a moving window type fingerprint verification method and verification device that prevent verification failures due to failure in alignment of one registration window.

[0020]

[Means for Solving the Problems] In order to achieve the above object, the present invention includes a plurality of alignment windows, and performs pattern matching by simultaneously scanning the plurality of alignment windows with respect to an input fingerprint image. The amount of movement of the input fingerprint image with respect to the registered dictionary is grasped, and pattern matching of the comparison window is performed.

Therefore, as shown in FIG. 1(A), a plurality of alignment window images WA, WB, WC (three in this example) are created.
and matching window images W1, W2, W3, W4 (4 in this example)
) will be established.

[0022]

[Operation] When performing verification, as shown in FIG. 1(B), a plurality of alignment window images WA, WB, and WC are simultaneously moved and aligned with the verification image. When alignment is established, the matching window image is then moved based on movement information based on this alignment, and matching is performed using the moving window method to verify the identity.

[0023] Since a plurality of alignment windows are provided in this way, even if one of them does not match due to a reason such as an injury to a finger, the alignment can be performed based on the other alignment window images. can.

[0024]

[Embodiment] Before describing the present invention in detail, the creation of a fingerprint dictionary will be explained with reference to FIG. ■When creating a fingerprint dictionary to be registered, first input the fingerprint image, and then
It is converted into a value and thinned to a 1-bit width.

[0025]Next, scanning is performed in a spiral pattern (either right or left) starting from the center to sequentially extract feature points.

[0026] After extracting a large number of feature points in this manner, the fingerprint image is cut out in a size of, for example, 16 x 16 pixels, centering on the feature points, and a specific number of points from the center are used as alignment windows. For example, in the example of FIG. 1(A), WA to WC
These three binary images are registered as alignment window images in a dictionary together with the position data.

[0027] Then, a predetermined number of feature points around the feature points are similarly cut out as matching windows, and these binary images are registered in the dictionary together with the position data as matching window images. Of course, personal information is also registered at this time. In addition, Figure 1
The example in (A) shows a case where four windows, W1 to W4, are used as matching windows.

When registering, in addition to extracting alignment windows obtained by scanning sequentially from the center as shown in FIG. 2, feature points located in close positions are grouped and extracted for alignment as shown in FIG. It can also be used as a window. This will be explained with reference to FIG.

■When creating a fingerprint dictionary to be registered, first input the registrant's fingerprint image, convert it into a binary value, and then
Thin lines to bit width.

[0030]Next, scanning is performed in a spiral manner (either right or left) starting from the center, and feature points are sequentially extracted.

[0031] After extracting a large number of feature points in this manner, a plurality of feature points that are located close to each other are extracted and grouped.

[0032] The fingerprint image is cut out in a size of, for example, 16 x 16 pixels, centering on the grouped feature points, and these groups are used as alignment windows. Then, these binary images are registered in the dictionary together with the position data.

[0033] Then, a specified number of other feature points are similarly cut out as matching window images, and these are registered in the dictionary together with the position data. Of course, personal information is also registered at this time.

In the case of FIG. 3, since the alignment window images are located close to each other, there is no need to move unnaturally large distances when verifying the identity. The effect of relative positional deviation of the windows can be reduced.

When extracting registered points in FIGS. 2 and 3, an example has been described in which scanning is performed in a spiral pattern from the center, but extraction of registered points is not limited to this spiral scanning; for example, in the horizontal direction Other means can be used, such as sequential scanning.

A block diagram of an embodiment of the present invention is shown in FIG. In FIG. 4, 1 is a fingerprint image input section, 2 is a fingerprint dictionary section, 3 is a fingerprint matching section, 4 is a matching number recording section, 5 is a position update section, 6 is a correction amount storage section, 10 is a control section, 11 is the keyboard.

[0037] The fingerprint image input unit 1 is used to input a fingerprint for registration or a fingerprint for comparison. At the time of registration, an alignment window WA is used as shown in FIG. ~Create a fingerprint dictionary image having WC and verification windows W1 to W4. Each of these images is entered in the fingerprint dictionary section 2 along with their position information, name, affiliation, etc. input from the keyboard 11, thereby forming a fingerprint dictionary. Naturally, the selection between fingerprint dictionary registration based on FIG. 2 and fingerprint dictionary registration based on FIG.

The fingerprint dictionary section 2 is composed of, for example, a memory, in which are entered the alignment window image, the verification window image, the position information of each of these window images, and personal information such as the name and affiliation of the registrant.

[0039] The fingerprint matching unit 3 performs fingerprint matching, and performs alignment using the alignment window image, and performs verification after alignment using the alignment window image. All of these are performed using a moving window.

[0040] The verification number storage section 4 records and holds information on matching of the alignment window image to the verification fingerprint in the fingerprint verification section 3, matching information on the verification window image, and the like.

The position updating section 5 controls the fingerprint matching section 3 to change the positions of the alignment window and the matching window for moving window control.

The correction amount storage section 6 stores the amount of movement of each alignment window and comparison window when a match is obtained by matching the alignment window image and the comparison window image.

[0043] The control unit 10 performs comprehensive control in fingerprint registration and fingerprint verification, and is composed of, for example, a CPU.

The keyboard 11 is used for manually inputting various data, such as inputting a selection instruction for dictionary registration as shown in FIG. 2 or 3, and inputting personal information such as the name and affiliation of the registrant.

Next, referring to FIG. 4, a case will be described in which the verification process shown in FIG. 1 is performed. At this time, the fingerprint dictionary unit 2 contains information (position, pattern, etc.) on the alignment window images WA to WC;
Information (position, pattern, etc.), personal information, etc. of the matching window images W1 to W4 are stored.

When a fingerprint image of a person seeking confirmation, that is, a person to be confirmed, is input to the fingerprint image input section 1, the fingerprint image input section 1 binarizes it and sends it to the fingerprint matching section 3. At this time, the alignment window image is transferred from the fingerprint dictionary section 2 to the fingerprint matching section 3.
Send to. For example, WA to WC in FIG. 1(A) are sent out.

[0047] The fingerprint matching unit 3 performs alignment by comparing these alignment window images WA to WC based on their position information. It is fine if the matching is successful, but if it is not successful,
The position update unit 5 simultaneously scans the moving windows of these alignment window images WA to WC and aligns them.

When the alignment is successful, the amount of movement at the time of successful alignment is stored in the correction amount storage section 6. Then, place the verification windows W1 to W4 at positions moved by this amount of movement,
Verify using sequential moving windows.

[0049] In this comparison, if the matching using the alignment window is unsuccessful, the next registered pattern is read out from the fingerprint dictionary section 2 and matching is performed using the alignment window. Even when it is successful, the next registered pattern is read out and verified through the alignment window. In this way, when a match is obtained with the fingerprint registered in the fingerprint dictionary section 2, the identity is verified, but if it does not match with all the registered fingerprints,
Identity verification is refused.

The method of determining whether or not the alignment is successful in the alignment window will be explained with reference to FIG. FIG. 5A shows a case in which it is determined whether or not the alignment was successful based on the number of matching windows. For example, when there are three alignment windows, if two match, it is determined that the alignment is successful. FIG. 5B shows a case where success is determined based on the proportion of all pixels that match in the alignment window.

The case of FIG. 5(A) will be briefly explained. (2) When a fingerprint image of a person to be confirmed is input to the fingerprint image input section 1, it is binarized and sent to the fingerprint matching section 3. At this time, the alignment image is sent from the fingerprint dictionary section 2 to the fingerprint matching section 3. The fingerprint matching unit 3 first performs alignment by comparing the position information of the alignment window image, but if they do not match, the entire alignment window image is moved at the same time to perform alignment. Now, if the degree of coincidence of the alignment windows is A, it is judged as a pass, and the degree of coincidence of each alignment window is A.
Determine whether or not it has been reached.

[0052] On the other hand, if it is determined whether the alignment is successful or not, for example, by whether half of all the alignment windows have reached the above A, then the alignment that has reached the matching degree A by the above moving The number of reference windows is monitored by the control unit 10, and when it reaches the specified number of windows even after moving within a predetermined range, the alignment is considered successful and the process proceeds to the next matching window. Read out the registered pattern and perform alignment verification.

The case of FIG. 5(B) will be explained. (2) When the fingerprint image of the person to be confirmed is input to the fingerprint image input unit 1, the alignment window image comparison is performed in the same manner as in the above case. At this time, the number of matches for each pixel in each alignment window is determined.

[0054] Then, when the degree of coincidence with respect to the total number of pixels of the entire alignment window reaches a specified value, it is judged as passing. For example, in WA to WC, if the matching degree of WA and WB is 80% and the matching degree of WC is 20%, and the matching rate for the entire pixel is 50% or more, it is considered as a pass (80 + 80
+20)÷3=60, and the alignment is considered successful. When alignment is successful, the process proceeds to a matching process using the next matching window, and when alignment is unsuccessful, the next registered pattern is read out and alignment and matching is performed.

[0055] In this way, even if one of the alignment windows does not match for some reason, it is possible to perform comparison using the comparison window.

Another embodiment of the present invention will be explained with reference to FIG. In the case of FIG. 6, a roughly matching point is first found using a plurality of alignment windows, and then it is determined whether this matching point is a correct alignment point.

(2) When a fingerprint image of the person to be confirmed is input to the fingerprint image input section 1, it is binarized and sent to the fingerprint matching section 3. At the same time, an alignment image is also sent from the fingerprint dictionary section 2.

[0058] The fingerprint collation unit 3 performs alignment by comparing the position information of this alignment window image, but at this time, the degree of coincidence is set to a slightly lower threshold. Then, this alignment window image is simultaneously moved using the moving window method to detect whether the degree of matching has been reached. At this time, even if all of the plurality of alignment windows do not reach the threshold, if they reach a predetermined value, such as a majority, the matching degree is determined to be equal to or greater than the threshold. Then, when the matching degree does not reach the threshold even after moving by a certain amount of movement, the alignment window of the next registered pattern is read out and collated.

(2) When the degree of matching reaches the threshold, change the threshold for pattern matching with respect to the alignment window image. At this time, the threshold value is changed to a value higher than the threshold value of (2) above.

■Then, pattern matching is performed on this alignment window using a moving window within a predetermined narrow tolerance range centered on the position where the degree of matching has reached the threshold in ■, and the high threshold determined in ■ is applied. Perform pattern matching.

[0061] As a result, when the number of alignment windows for which a degree of matching is obtained with a high threshold value does not reach a specified value, the identity verification is rejected and the next registered pattern is read out from the fingerprint verification unit 3.

[0062] When the number of alignment windows for which a degree of matching was obtained using a high threshold value reaches a specified value, next comparison is performed using a comparison window image.

According to the example shown in FIG. 6, a plurality of alignment windows are moved simultaneously to obtain an alignment candidate point using a lower threshold value, and then it is determined whether or not this candidate point is a correct alignment point using a higher threshold value. Therefore, it is possible to perform verification faster than by performing fine movements to judge the entire image using a high threshold.

[0064]

According to the present invention, even if one of the alignment windows cannot be aligned, the other alignment windows can be used to perform alignment, thereby improving the accuracy of matching.

Furthermore, even if a plurality of alignment windows are provided, the plurality of alignment windows are simultaneously compared with a low matching degree at first;
When positioning point candidates are obtained, accurate positioning can be performed using each positioning window, so even if a plurality of positioning windows are provided, positioning can be performed quickly.

[Brief explanation of drawings]

FIG. 1 is an explanatory diagram of an embodiment of the present invention.

FIG. 2 is an explanatory diagram for creating a dictionary used in the present invention.

FIG. 3 is another diagram for explaining the creation of a dictionary used in the present invention.

FIG. 4 is a block diagram of the configuration of an embodiment of the present invention.

FIG. 5 is an explanatory diagram of the alignment state in the present invention.

FIG. 6 is an explanatory diagram of a second embodiment of the present invention.

[Figure 7] (A) is a fingerprint ridge, (B) is a fingerprint dictionary image, (
C) shows a matching state.

FIG. 8 is an explanatory diagram of a conventional collation operation.

[Explanation of symbols]

1 Fingerprint image input section 2 Fingerprint dictionary section 3 Fingerprint verification section 4 Verification number recording section 5 Location update section 6 Correction amount storage section 10 Control section 11 Keyboard

Claims (6)

[Claims] Claim 1: A moving window fingerprint image matching method comprising a fingerprint registration dictionary having a plurality of minutiae window images of registered fingerprints and matching an input image with these plurality of window images. Equipped with multiple window images,
These multiple alignment windows were scanned simultaneously against the input fingerprint image, pattern matching was performed, the amount of movement of the input fingerprint image relative to the registered dictionary was determined, and the dependent matching window was moved by this amount of movement. A moving window fingerprint image matching method characterized in that pattern matching of matching windows is then performed. [Claim 2] When determining whether a plurality of alignment windows match, the coincidence or mismatch of each alignment window is determined, and when the number of matching alignment windows is greater than or equal to a predetermined threshold, the position is determined to be correct. 2. The moving window type fingerprint image matching method according to claim 1, wherein the moving window type fingerprint image matching method is characterized in that it is determined that the fingerprint images are matched. [Claim 3] When determining whether a plurality of alignment windows match, correct alignment is determined when a pixel that matches a predetermined threshold or more is detected with respect to the pixel sum of all used alignment window images. Moving window -type fingerprint imaging method according to claim 1, which is characterized by distinguishing as a thing. 4. When determining the coincidence of a plurality of alignment windows, first, the relative positional relationship of the plurality of alignment window images is left as is and they are simultaneously moved, and the degree of coincidence is determined using a first threshold value. Then, move the alignment window image to
4. The moving window fingerprint image matching method according to claim 1, wherein the degree of matching is determined using a second threshold value higher than the threshold value of . 5. Fingerprint image input means into which a fingerprint image to be compared is input; a fingerprint registration dictionary having a plurality of minutiae window images of registered fingerprints; and a fingerprint registration dictionary having a plurality of minutia window images of registered fingerprints; A moving window type fingerprint image matching device for matching images includes a plurality of alignment window images, a fingerprint dictionary means (2) having the matching window images, an input fingerprint image, and a fingerprint dictionary means (2). a fingerprint matching means (3) for matching the window image held in the fingerprint matching means (3); and a position updating means (5) for controlling the movement of the image in the fingerprint matching means (3).
A moving window type fingerprint image matching device, comprising: comparing an input fingerprint image with a plurality of alignment window images registered in a fingerprint dictionary, and then comparing the input fingerprint image with the matching window images. 6. The moving window according to claim 5, further comprising fingerprint dictionary means storing a plurality of feature point window images having a close positional relationship as the plurality of alignment window images. type fingerprint image matching device.