JPH01271882A - Collation system for plastic partial picture - Google Patents

Abstract

PURPOSE:To back up the weak point of each collating method and to realize a highly reliable personal collation system by defining the AND between two collation results of a pattern matching method and the comparison of the local structure features as the success conditions. CONSTITUTION:At collation of the personal feature information, a feature point partial picture 11 is called out of a registered personal feature information memory part 3. Then the collation of the personal feature information is carried out by a positioning part 6 and a minor area picture collating part 7 based on the processing procedure of a normal plastic partial picture collating method. Then the degree of coincidence is checked among the feature point partial pictures for each feature minor area after a positioning job. For the successful minor areas having the degree of discordance less than a prescribed threshold level, these minor areas are really turned into thin lines by a minor area feature collating part 8 for checking the presence or absence of the feature points within the minor areas. If the minor area contains a feature point, it is checked whether the type of the feature point is coincident or not with a feature code 13 read out of the part 3. Thus the success is finally decided only when both a binarized image and a feature code are successful.

Description

[Detailed Description of the Invention] [Summary] The present invention relates to a plastic partial image matching method, and aims to achieve a similar discrimination ability regardless of the type of fingerprint ridge pattern. This method uses a combination of pattern matching method and local structural feature extraction/analysis for small regions that have feature points, and generates an image of the small region that includes feature points, that is, a feature point partial image, and this feature point partial image. The feature code indicating the feature and the coordinates of the feature small area are registered for each individual, and for personal identification, after positioning, the above feature point partial image is scanned over the input fingerprint image, and Regarding a small area that is determined to match the feature point partial image within the range of , the feature code corresponding to the feature of this small area is checked with the feature code of the feature point partial image corresponding to the small area. By doing this, it is possible to prevent the occurrence of false matching in plastic partial image matching.

[Industrial application field]

The present invention relates to a plastic partial image matching method, and more particularly to a highly reliable fingerprint matching method that combines matching of partial images and reconfirmation using feature codes of the partial images.

[Conventional technology]

Currently, fingerprint recognition technology has become an extremely important branch of the field of personal identification, and further improvements in recognition performance are desired.

In recent years, as computers have been widely introduced into social systems, people concerned have focused on how to ensure system security. For example, many questions have been raised regarding the security of ID cards and passwords, which have been used until now as a means of verifying identity when entering a computer room or using a terminal. Fingerprints, on the other hand, are “the same for everyone”°,“
Because fingerprints have three major characteristics: ``unchangeable throughout life,'' they are considered the most powerful means of identity verification, and much research and development is being conducted on simple personal identification systems using fingerprints.

As a conventional means for fingerprint image matching, a plurality of characteristic partial images are extracted from the entire reference fingerprint image, and these are registered in advance in an external medium as minutiae partial images for each individual. There is a known method in which pattern matching is performed by scanning these on an input fingerprint image.

Furthermore, there are two ways to perform pattern matching: scanning a fingerprint image while completely constraining (fixing) the relative positional relationships of multiple minutiae partial images, and scanning each minutiae partial image by removing the constraints. There are some that can be scanned freely.

If the relative positional relationships of the feature point partial images are constrained during the pattern matching described above, it will not be determined as a "match" unless they completely match the original relationship, so differences in fingerprint pressing, etc. When distortion occurs in a fingerprint image due to the above, it has not been possible to deal with the distortion.

On the other hand, the method of freely scanning each feature point partial image can deal with the above distortion, but the pattern matching may occur at a position that is completely different from the original relative position of each feature point partial image. , even if it is someone else's fingerprint, it is mistakenly "matched"
There was a danger of making a judgment.

Therefore, the present inventors previously proposed a fingerprint image matching method in Japanese Patent Application No. 61-220-5=869 that deals with the distortion of fingerprint images and enables pattern matching without mismatching. This is a fingerprint matching method in which an input fingerprint image is matched with a plurality of minutiae partial images stored in advance for fingerprint verification, and a match or mismatch between the images is determined. One of the feature point partial images is scanned on the input fingerprint image, the degree of mismatch between the images is calculated for each position, and the one feature point is placed at the position where the degree of mismatch is the lowest. After moving the partial image and parallelly moving the remaining feature point partial images accordingly, the remaining feature point partial images are scanned in the vicinity of the position after the parallel movement, and each partial image is Find the degree of discrepancy between the images,
The matching is determined based on whether a position showing a degree of inconsistency less than or equal to a predetermined darkness value is found.

By determining the mismatch, the distortion of the fingerprint image can be dealt with and the occurrence of erroneous matching can be prevented.

The main configuration of a personal verification device for identity verification is based on the number of matching pixels between the above-mentioned personal characteristic information (for example, information unique to an individual such as a fingerprint or a retinal image) and a pre-registered feature point partial image. Shown in Figure 4.

In the figure, 1 is an input section for personal characteristic information, which corresponds to a fingerprint sensor 9'' when a fingerprint is used as the personal characteristic information.2 is a personal characteristic information-time storage section, which is detected by the fingerprint sensor. A frame J is a memory that temporarily stores a binary image of a fingerprint image. 3 is a registered personal feature information storage unit that stores a plurality of feature point partial images registered in advance for each individual and their registration time. This is a dictionary that stores the coordinates of the person.It is a dictionary that stores the coordinates of the person's personal identification card.It is a dictionary that stores the coordinates of It may be something.4
is a personal characteristic information comparison section, which is a registered personal characteristic information storage section (3 to 7), and a positioning section which positions the extracted feature point partial image on the input fingerprint image quality. It is equipped with a small area 1.5 mark matching unit that performs a small area image of a fingerprint image and a 'l, +I feature point partial pixel partial image.

If the above-mentioned plastic partial image matching method is implemented using this personal matching device, the number of matching pixels is smaller compared to the full-scale matching method, so there is an advantage that the speed can be increased.

[Problems to be Solved by the Invention] However, by comparing the corresponding pixels of both the registered personal characteristic information and the small area image cut out from the input personal characteristic information, it is possible to compare the corresponding pixels of both the registered personal characteristic information and the inputted personal characteristic information, and to determine whether the number of mismatched pixels is below a predetermined darkness value. Sometimes, when it is judged as a "match", if there are many parallel ridges as in the fingerprint shown in Fig. 5(a), "end points" as shown in Fig. 5(b) are used.
There is a risk that a feature point partial image containing ``matches'' with an originally separate image as shown in FIG.

Figure (a) shows a fingerprint image detected by a fingerprint sensor and output after image processing by a computer. FIG. 2B shows an enlarged view of a small region surrounded by a thick line. As described above, the conventional plastic partial image matching ability has a problem in that the discrimination ability varies depending on the pattern of the ridges of the fingerprint.

Therefore, it is an object of the present invention to make it possible to obtain various discrimination abilities related to the type of pattern of fingerprint ridges.

[Failure to solve the problem]

In the present invention, as personal characteristic information registered in advance for verification, a feature code indicating the characteristics of each feature point partial image is added to the coordinates of the feature point partial image and the small area, and a feature code is added to the coordinates of the feature point partial image and the small area. A small area feature matching section is added to the personal feature information comparing section.

When performing personal verification, the above-mentioned plastic partial image matching method is applied to an input personal characteristic information image such as a fingerprint image using a plurality of feature point partial images of the registered personal characteristic information. When executing the process, a pattern matching method is used to compare local structural features between the feature point partial image that has been determined to be successful and the small region image on the fingerprint image. We re-examined the images used and determined that they matched only when the images matched and the structural features matched.

-9= [Function] The present invention uses two matching methods: a pattern matching method that compares the mismatch degree of pixel information in a small area with a predetermined darkness value, and a comparison of local structural features. By taking advantage of the fact that they are different and their strengths and weaknesses complement each other,
By setting the logical product of the memory verification results as a passing condition, the weaknesses of each verification method are compensated for and highly reliable personal verification is realized.

(Embodiment) An embodiment of the present invention will be described below with reference to Figs. 1 to 3. In this embodiment, an example in which a fingerprint is used as personal characteristic information will be explained.

FIG. 1 is a diagram for explaining the configuration of the main parts of the personal verification device used in this example, in which 1 is a personal characteristic information input section;
It is a sensor that detects a pattern unique to each individual used in personal verification, and in this embodiment, which uses a fingerprint as personal characteristic information, the fingerprint sensor → ノ, 2 is the personal characteristic information-time storage unit, which is collected by the fingerprint sensor. a frame memory for temporarily storing the fingerprint image that has been binarized and output; 3;
is a registered personal characteristic information storage unit, the configuration of which will be described later.

Reference numeral 4 denotes a personal characteristic information comparison section, which has conventionally included a positioning section 6 and a small area image matching section 7 as its constituent elements, but this embodiment further includes a small area feature matching section 8.

The registered personal characteristic information storage section 3 may be provided in a personal verification device as described above, or may be a magnetic card, an IC card, etc.
It may also be a portable storage device such as a card. As shown in Fig. 2, the contents include a conventional minutiae partial image 11, which is a binarized pattern of an image of a small area that has minutiae inside, and the coordinates of the minutiae in the fingerprint pattern taken at the time of registration. A feature code 13 indicating the type of feature of the feature point partial image 11 is added to the feature small area coordinates 12.

Next, the registration and verification operations of this embodiment will be explained.

When registering a fingerprint as personal characteristic information, the personal characteristic information input unit (hereinafter abbreviated as fingerprint sensor) 1 digitalizes and thins a manually generated fingerprint image.

This fingerprint image is divided into small regions of a predetermined size, and feature points such as end points and branch points are determined for each small region.

Next, small regions having feature points are found in order from the small region closest to the center of the visual field, and the feature small regions are cut out so that the feature points are located approximately at the center of the small region. For each feature small region, a feature point partial image 11. Feature small area coordinates 1
2. and the type of feature points existing in each feature small region [third
See diagram.

The details will be described later] is stored in the registered personal characteristic storage section 3. As mentioned above, this registration information may be recorded in a storage device installed in the personal verification device.
It is also possible to record on a portable storage medium.

When collating personal characteristic information, first, the focal point partial image 11 is called from the registered personal characteristic information storage section 3.

After that, matching is performed by the alignment unit 6 and the small area image matching unit 7 according to the processing procedure of the normal plastic partial image matching method, and after alignment, the matching degree of the feature point partial images is calculated for each feature small area. investigate. For a passing small region whose mismatch degree is less than a predetermined dark value, the small region feature matching unit 8 further thins the line and checks whether there are feature points in the small region, and if the small region has feature points, It is checked whether the type of feature point matches the feature code 13 read from the registered personal information storage section 3.

There are many types of feature point shapes as shown in Figure 3, but they are broadly classified into end points and branch points, and depending on their orientation, end points can be divided into four types, for example (a) to (d). , branch points can be classified into eight types, for example (e) to (1). Therefore, in (a) to (d), E-1 to E-4° (
Feature codes such as B-1 to B-8 can be assigned to e) to (1), respectively.

Such a code is stored as the feature code 13, and when comparing feature codes, the type of feature point in the feature small area that matches the feature point partial image is checked, and the feature code and the read feature code are Compare with 13.

In this embodiment, when the binarized images match as described above, it is further checked whether the feature codes match, and only when both pass, it is determined that the test is passed.

However, the pass condition is not to pass when both the binarized image and the feature code match for all small areas as described above, but to pass if the feature code matches among the small areas where the binarized image matches. If the percentage of those that meet the criteria exceeds a predetermined value, the test can be passed, and the pass judgment conditions can be modified in various ways. For example, it is possible to set the pass condition to be that the number of matching small areas for each type of feature code is greater than a certain percentage of the number of small areas categorized by that feature code, and also for the type of feature code. The above ratio can also be made different for each case. Alternatively, out of all the small areas in which the binarized images match, more than a percentage of the small areas may also match the feature code and pass the test. Depending on the rate of correct determination or, conversely, the risk of erroneously determining that the person in question is not the person in question, various choices can be made depending on the acceptance rate.

In this way, in this embodiment, the ambiguity that cannot be avoided when matching is performed using only one matching method is reinforced by using two matching methods with different matching principles, and the matching accuracy is improved by using two matching methods with different matching principles. This is what I did.

In other words, the matching is based on a pattern matching method that compares "1" and "O" of each pixel constituting a small region image with a feature point partial image, and compares the degree of mismatch of pixel information between the two with a predetermined darkness value. In the case of a fingerprint with many parallel ridges as shown in FIG. 5(a), the small area image of the L eel shown in FIG. 5(C) is converted to the feature points shown in FIG. It is difficult to distinguish it from a partial image.

Therefore, in the normal partial image matching method, there is a risk of erroneously determining the two stored images as a match.

However, since the small area image shown in FIG. 2C has no end points or branch points, there are no feature points inside. Therefore, by comparing the feature codes and extracting and analyzing local structural features, it can be easily detected that the two patterns are different types.

In this way, in this embodiment, two matching methods with different principles, the so-called pattern munching method and local structural feature comparison, are used together, and the only passing condition is when both of them are determined to be a match. It is possible to complement the ambiguity of matching that exists in each image matching method and improve matching accuracy.

Note that the present invention is not limited to the above embodiment,
It can be implemented with various modifications. That is, as the personal characteristic information used for personal verification, in addition to fingerprints, it is also possible to use, for example, a retinal pattern.

The fingerprint sensor used in the personal characteristic information input section may be a flat sensor using a hologram, a prism type fingerprint sensor, or the like, and can be selected depending on the personal characteristic information to be used.

Furthermore, the method of matching the small area of the input personal feature information image-L with the registered feature point partial image does not need to be particularly limited, and any commonly used partial image matching method may be used. . Furthermore, the positioning method is not limited to the one embodiment described above, and can be implemented with various modifications.

[Effects of the Invention] As described above, according to the present invention, by making it possible to improve the accuracy of matching by making it a passing condition that the degree of matching of binarized images also matches the feature code. did it.

[Brief explanation of the drawing]

Figure 1 is an explanatory diagram of the main part configuration of the personal verification device used in one embodiment of the present invention, Figure 2 is an explanatory diagram of the dictionary configuration of the above embodiment, and Figure 3 (a)
~(1) is an explanatory diagram of the coat proportion of the characteristic small area of the above-mentioned embodiment, FIG. 4 is an explanatory diagram of the configuration of a conventional personal verification device, and FIG. 5 is an explanatory diagram of problems in the conventional partial image verification method. Figure (a) shows an example of a binary image of a fingerprint taken by a fingerprint sensor and output after being processed by a computer. Figure (b) is an enlarged view of the small area surrounded by the thick line in (a). , the same figure (C) is the above (b)
It is a figure which shows the example which is easy to be erroneously collated. In the figure, 1 is a personal characteristic information input section, 2 is a personal characteristic information time storage section, 3 is a registered personal characteristic information storage section, 4 is a personal characteristic information comparison section, 5 is a judgment result, and 6 is a misaligned position. 7 is a small area image matching unit, 8 is a small area feature matching unit, 11
is a feature point partial image, 12 is a special point partial image,
J14: (Category ()IV) 116λ/11/11 1.1 Kangyu's 1988 Patent Application No. 154134 2 Name of the invention Plastic partial image matching method 3, relationship with the person making the amendment case Patent applicant address: Nakahara-ku, Yozaki City, Kanagawa Prefecture - Hi Oda [January 015 Street name (522) Fujitsu Ltd. 5, date of amendment order (shipment date) 815, 1985 (
a) (b) (c,)
(d) (e) (f)
(g) (h) (i)
01 (k) (1) Features of one embodiment of the present invention Code north side explanation of small area Same No. 3
figure

Claims (2)

[Claims] (1) A personal characteristic information input unit (1) that detects personal characteristic information to be verified; each of a plurality of partial images constituting registered personal characteristic information stored in advance for each individual; and the personal characteristic information. an alignment unit (6) that performs alignment with the personal characteristic information to be matched input from the input unit; and a positioning unit (6) that scans each of the plurality of partial images in the vicinity of a candidate position on the personal characteristic information to be matched, which is input from the input unit. and a personal characteristic information comparison unit (4) that determines the match or mismatch depending on whether or not a position showing a predetermined degree of mismatch is found for each partial image. For each of a plurality of small regions that have feature points inside as completed personal feature information, a feature point partial image (11) showing an image of the small region and feature small region coordinates (12) showing the coordinates of the small region. and a feature code (13) indicating the type of feature point in the small area, and a small area feature matching unit (8) in the personal feature information comparing unit (4).
The small area feature matching unit (7) performs the small area feature matching on the image of the small area in the personal feature information to be matched that is determined to match the corresponding feature point partial image (11). A plastic partial image matching method characterized in that the feature is checked by the section (8) and the feature code indicating the feature is compared with the feature code (13) of the corresponding feature point partial image (11). (2) The plastic partial image matching method according to claim 1, characterized in that a fingerprint is used as the personal characteristic information.