JPH04105168A - Individual cirtifying device - Google Patents

Abstract

PURPOSE:To improve the accuracy of collating and the reliability of cirtification and to reduce the generation of an error to exclude the person himself by registering feature information extracted from the finger of the different hands respectively in advance. CONSTITUTION:A room entering/leaving controller consists of a prism 21, a light source 22, a CCD camera 23, an A/D converter 24, a control part 25, an IC card 28, a reading/writing device 29, a door 30, a door driving part 31 and a memory 32. Here, since it can be compared with plural pieces of feature information for being collated extracted from the some finger pictures of which a condition is different little by little, the little fluctuation of the feature information owing to the difference of the way of setting of a finger F can be absorbed. Thus, the accuracy of collating and the reliability of cirtification can be improved and the generation of an error to exclude the person himself can be reduced.

Description

DETAILED DESCRIPTION OF THE INVENTION [Purpose of the Invention (Field of Industrial Application) The present invention relates to a personal authentication device used in a room entry/exit control device that manages people entering and leaving a room based on physical characteristics such as fingerprints.

(Prior Art) In recent years, entry/exit control devices that use the physical characteristics of individuals to control entry and exit from specific areas, for example, use finger image information in the longitudinal direction of the finger as an authentication method. A method has been developed that uses feature information obtained by adding in a direction perpendicular to .

FIG. 8 shows an outline of the room entry/exit control device according to the above-described method.

That is, the finger F placed on the finger rest 1 is illuminated by the illumination lamp 2.
The reflected light image is captured by the TV camera 3.
By photographing the finger F, image information of the finger F can be obtained. This image information is converted into a digital quantity by the A/D converter 4 and then sent to the control section 5. The pixel values of the image information sent to the control unit 5 are added by the feature extraction unit 6 in a direction orthogonal to the longitudinal direction of the finger F, thereby extracting feature information.

At this time, the characteristic information (characteristic information to be verified) requested by the person to be authenticated as described above has already been registered in the IC card 7 via the reading/writing device 8 .

The feature information extracted by the feature extraction section 6 is then read out from the IC card 7 in the comparison section 9 by inputting, for example, a pre-given ID number, and stored in the memory 1o. It is compared and verified with the feature information to be verified.

If the two match sufficiently, the control unit 5 determines that the person to be authenticated is the person to be authenticated, and the drive unit 1 opens the door 12. On the other hand, if no match is found, it is determined that the person is a stranger, and the door 12 will not be opened.

By the way, the problem with this type of access control device is that
This is a case where distortion occurs in the characteristic information due to a difference in the way the finger F is placed. In general, the feature information extracted from the input finger image is not necessarily the same every time, but usually involves some variation. Therefore, if a sufficient match is not found through comparison and matching, even if the person is the person himself/herself, if there is a large variation, an error may occur in which the person is determined to be someone else (so-called Type 1 error). was there.

As a solution to this problem, a method has been considered in which the finger image is continuously inputted while the finger F is placed on the finger rest 1, and each time the finger image is compared and verified with the feature information to be matched. .

However, in this method, finger images with slightly different states are input, and as a result of several inputs, it is hoped that feature information that closely matches the pre-registered feature information to be matched will be obtained. be. For this reason, it is effective when the feature information to be matched is in a standard state, but if the state of the feature information to be matched is poor, it is not necessarily possible to obtain a good result from a newly input finger image. The drawback is that it is not always possible to extract matching feature information.

(Problems to be Solved by the Invention) As described above, conventionally, finger images are continuously input while the finger is placed on the finger rest, and each time,
Even if comparison and matching with the feature information to be matched is performed, if the condition of the feature information to be matched is poor,
Since it is not always possible to extract characteristic information that matches well from newly input finger images, the accuracy of matching is poor and the reliability of authentication is poor, such as being excluded as someone else even though the person is the real person. It had the disadvantage of being low.

SUMMARY OF THE INVENTION Therefore, an object of the present invention is to provide a personal authentication device that can improve the accuracy of verification and the reliability of authentication, and can reduce the occurrence of errors that exclude the person in question.

[Structure of the Invention] (Means for Solving the Problems) In order to achieve the above object, the personal authentication device of the present invention includes at least the first. an input means for inputting image information of a finger including a second joint; a feature extraction means for extracting feature information from the image information of the finger input by the input means; and an input operation by the input means and the feature extraction means A feature registration means for registering a plurality of pieces of feature information obtained by repeating the extraction operation by repeating the extraction operation as feature information to be matched, and a plurality of pieces of feature information to be matched registered in advance in this feature registration means; a comparison means that continuously compares the feature information extracted by the feature extraction means from the finger image information input by the input means;
It is comprised of a determining means for determining whether or not the feature information and the feature information to be matched are extracted from the same finger based on the comparison result by the comparing means.

Further, in the personal authentication device of the present invention, at least the first. an input means for inputting image information of a finger including a second joint; a feature extraction means for extracting feature information from the image information of the finger input by the input means; and an input operation by the input means and the feature extraction means a feature registration means for registering each feature information obtained from a plurality of fingers by repeated extraction operations as feature information to be matched; a plurality of feature information to be matched registered in advance in the feature registration means; Separately, a comparing means continuously compares the feature information extracted by the feature extracting means from the image information of the finger inputted by the input means, and based on the comparison result by this comparing means, determination means for determining whether or not there is any matching with the feature information in the feature information for use, and if the determination means does not determine a match, the comparison operation by the comparison means is replaced with another finger. The structure is such that it is carried out by

Further, in the personal authentication device of the present invention, at least the first. an input means for inputting image information of a finger including a second joint; a feature extraction means for extracting feature information from the image information of the finger input by the input means; and an input operation by the input means and the feature extraction means a feature registration means for selecting feature information having low correlation with each other from a plurality of feature information obtained by repeated extraction operations and registering this as feature information to be matched; and a plurality of objects registered in advance in this feature registration means. a comparing means for continuously comparing the matching feature information and the feature information extracted by the feature extracting means from the finger image information separately input by the input means; and a comparison result by the comparing means. and determining means for determining whether or not the feature information and the feature information to be matched are extracted from the same finger based on the same finger.

Furthermore, in the personal authentication device of the present invention, at least the first. an input means for inputting image information of a finger including a second joint; a feature extraction means for extracting feature information from the image information of the finger input by the input means; and an input operation by the input means and the feature extraction means a feature registration means for registering each feature information obtained from a plurality of fingers as feature information to be matched by repeating the extraction operation by;
Comparison of a plurality of pieces of feature information to be matched registered in advance in this feature registration means and feature information extracted by the feature extracting means from finger image information input separately from the input means. and authenticating an individual based on the comparison result of the comparison means, in which at least one of the plurality of feature information for verification registered in the feature registration means is used by another hand. It is said that the composition is determined from the fingers.

(Function) With the above-described means, the present invention enables comparison with a plurality of pieces of feature information to be matched extracted from several finger images with slightly different states. This makes it possible to absorb slight variations in feature information.

(Example) Hereinafter, an example of the present invention will be described with reference to the drawings.

FIG. 1 shows an example of a personal authentication device of the present invention, such as an entry/exit control device for controlling entry to an important area.

In FIG. 1, a light source 22 that irradiates light from inside the surface of the prism 21 (finger rest) is provided near the prism 21 as a finger rest. For example, a CCD camera 23 is provided which receives reflected light from the finger F and captures image information of the finger including at least the fingerprint and the first and second joints.

In this case, depending on whether the finger F and the prism 21 are in contact or non-contact, the light from the light source 22 is totally reflected at a concave portion such as a fingerprint, and is diffused at other convex portions.

Therefore, by receiving part of the light diffused by the convex portion of the finger F as reflected light, image information of the finger can be captured.

The image information of the finger output from this camera 23 is
(analog/digital) converter 24.

The finger image information supplied to the A/D converter 24 is converted into a digital image signal IM(X, )') of, for example, 512×512 pixels and 8 bits/pixel.

1M(x,y) (x-y-0,1,2,...,511) ...(
1) The finger image signal IM (x, y) output from this A/D converter 24 is supplied to the control section 25.

The finger image signal IM (x, y) sent to the control unit 25 is
The feature extraction unit 26 adds the pixel density in a direction perpendicular to the longitudinal direction of the finger F, thereby obtaining feature information P (x).
is extracted.

The feature information P(X) extracted by this feature extraction unit 26 is
In the comparison unit 27, the characteristic information for verification obtained from the finger F of the person to be authenticated registered in advance in the IC card (described later) (this is obtained in the same manner as the characteristic information P (x) described above) is a one-dimensional summed signal).

Here, a reading/writing device 29 that exchanges information with, for example, an IC card 28, a door drive section 31 that controls opening and closing of the door 30, and a memory 32 are connected to the control section 25, respectively.

The IC card 28 has a storage section, and is configured to store in advance various information and a plurality of sets of characteristic information to be verified for the person to be authenticated as a registered image.

Each area 32a, 32b, . . . 32n of the memory 32 is configured to store a plurality of sets of feature information to be verified, which are read from the IC card 28 at the time of verification.

The door 30 is opened and closed only when the control section 25 determines that the person to be authenticated is the authenticated person based on the comparison result by the comparison section 27.

FIG. 2 schematically shows the relationship between a finger image (finger image signal) and feature information.

The finger image IM(X, )/) is an image pattern of the finger F read by the camera 23 and digitized by the A/D converter 24.

The feature information P (x) is expressed as the finger image IM (x) in the direction y perpendicular to the longitudinal direction X of the finger F.

This is a one-dimensional signal obtained by adding the pixel densities of y). This characteristic information P (x) has steep valleys (minimum values) at the positions of horizontal wrinkles corresponding to each joint, and this serves as a parameter (individuality) for authenticating an individual.

Next, the operation in the above configuration will be explained.

FIG. 3 shows the procedure for registering finger characteristic information. Here, it is assumed that n pieces (multiple sets) of characteristic information are registered for a certain finger (one finger) of each person to be authenticated.

First, i is set to 1 as an initial value indicating the number of registrations (step 5TI).

After that, the light source 2 is placed on the finger F that is in contact with the surface of the prism 21.
2 is irradiated with light, and a reflected light image is photographed by a camera 23. Then, the output is converted into a digital signal by the A/D converter 24, so that at least the fingerprint and the first . Finger image IM (x, y) including the second joint part
is input (step 5T2).

Next, the finger image IM('x, y) is sent to the feature extraction section 26 in the control section 25, where feature information P (x) is extracted as described above (step 5T3). In this case, since it is possible to register n times of characteristic information P (x) for each person to be authenticated, each characteristic information P (x) is expressed as follows.

Pot(X) (x = 0.1, 2.-.511, i = 1.
2, 3. - , n) - (3) Therefore, the feature information Po+(x) for the input of the first finger image IM (x, y) obtained by the feature extraction unit 26 is determined by the control of the control unit 25. The information is stored in the IC card 28 via the reading/writing device 29 (step 5T4). As a result, the feature information Po
t(x) is the IC card 28 as feature information 1 to be verified.
will be registered.

Feature information P for the first input of finger image IM (x, y). When the registration of 1(X) is completed, the feature information P for the second and subsequent inputs of finger images IM (x, y) is
. , (x) are registered in the same way. That is, each time the registration is completed, i, which is an initial value indicating the number of registrations, is counted up by one (step 5T5).
). Then, the steps ST2, 3, and 4.5 described above are
It is repeated until a predetermined number n is reached (step 5T6)
.

As a result, feature information P for the input of each finger image IM (x, y). 2(x) to POfi(X) are registered as feature information to be matched 2 to n, respectively.

Note that the input of the finger image IM (X, y) here is not performed continuously by keeping the finger F on the prism 21, but by repositioning the finger F one at a time and inputting each time. It's good. Also, rather than performing this registration in a short period of time,
This may be performed at sufficient time intervals.

FIG. 4 shows the operation when, for example, when entering and leaving a room, the characteristic information of a finger is verified in a state where a plurality of sets of characteristic information to be verified are registered in the IC card 28 as described above. .

For example, assume that the IC card 28 in which characteristic information of an individual's finger is registered is set in the reading/writing device 29 by the person to be authenticated. Then, i as an initial value is set to 1 (step 5T11). Then, until this initial value i reaches a predetermined number n, n pieces of characteristic information 1 to n to be verified registered in the IC card 28 are sequentially read out and loaded into each area 32a to 32n of the memory 32, respectively. (Step 5T12, 13.14). It is assumed that these feature information items 1 to n to be verified are expressed by the above-mentioned equation (3).

In this way, when the loading of the feature information 1 to n to be verified is completed, j as an initial value indicating the number of verifications is set to 1 (step 5T15).

On the other hand, the person to be authenticated places a predetermined finger F on the finger placement surface of the prism 21.
, input the finger image IM(x,y) (step 5T1
6) and extraction of feature information P (x) (step 5T17)
is repeated.

Here, in this embodiment, m
Since the verification based on the input of the finger images IM(x, y) is performed continuously with the finger F kept on the prism 21, for example, the comparison of the finger images obtained by the fifth input is performed continuously. The feature information P (x) is expressed as follows.

PRI(X) (x = 0.1, 2..., 511, j -1, 2
,3,...,m)...(4) Also, one finger image I
M (x, y) input (feature information PRt extracted from this
(X)+, n pieces of characteristic information to be compared 1 to n, that is, the characteristic information P registered by each ninja witness. , (X). Here, the matching (well, positioning is performed while calculating the square error between the feature information PRI(X) and the feature information P.1(x). Let the verification value be S.

s-v (Pot (X), PR1(X))
-(5) That is, the squared error f (t) at a certain position t is obtained by f (t) (6). Therefore, the matching value S at this time is S-min (f (t))
−(7). However, t is −tl<
The range is t<tl.

The smaller the matching value S is, the more the two feature information P
. , (x), and PRI (X) are very similar.

Further, whether or not the person is the real person is determined based on the magnitude relationship between the verification value S obtained above and a threshold value T, which is a criterion for determination, which has been obtained, for example, experimentally. That is, when S ≦ T... When the person himself/herself extracts the feature information PRI (X) for the first input finger image IM (x, y), the initial value i
is set to 1 again (step 5T18), and
This feature information PRI(X) is first compared with feature information 1 to be matched represented by Pot(X) (step 5
T19).

As a result of this matching, the matching value S obtained is the condition (S≦T)
If it satisfies (step 5T20), it is determined that the person is the real person (step 5T21). Then, the control unit 25 unlocks the door 30, and the door drive unit 31 opens the door 30, allowing the person to be authenticated to enter the room (step 5T22).

On the other hand, in step 5T20, if the obtained verification value S does not satisfy the above conditions, the initial value i is incremented by one (step 5T2B).
, the feature information PRI(X) and the feature information to be matched 2 are compared in the same manner.

In this way, the first finger image IM(x, y
) The feature information PR□(x) for the input of
T19, 20.23 are repeated. Note that during this process, if the verification result satisfies the above conditions, step 5T21.
After moving to step 22, the process ends.

Further, if a result satisfying the above conditions is not obtained even after matching with n pieces of feature information to be matched 1 to n, the initial value j is incremented by one (step 5T25). Then, by repeating the processes from step 5T16 described above, the second finger image IM (
The above-mentioned matching process for the inputs (x, y) is performed in the same manner.

Similarly, until the initial value j or the predetermined number m is reached (step 5
T26), if a result satisfying the above conditions is not obtained even after repeating the processes from step 5T16 onward, it is determined that the person to be authenticated is a stranger (step 5T2).
7), the process is terminated.

Note that the input of finger images is performed continuously while the person to be authenticated keeps their finger on the prism, so the person to be authenticated should input the finger little by little so that the state of the finger image changes slightly. It is more effective to move the

As described above, it is possible to compare a plurality of pieces of feature information to be matched extracted from several finger images with slightly different states.

That is, the finger image IM(x) input at the time of verification.

The feature information PRI(X) extracted from y) is sequentially matched with n pieces of feature information to be matched 1 to n.

This makes it possible to absorb slight variations in feature information due to differences in finger placement, etc. I want to, 1
Even if a match is not recognized by matching with the first feature information to be matched, if a match is found by matching with the second feature information to be matched, the identity of the person can be determined. This can reduce errors such as erroneously determining that a person is someone else and excluding the person.

Additionally, since the finger image input during verification is performed continuously while the person being authenticated keeps their finger on the prism, it is possible to obtain an image that closely matches the feature information for the person to be verified. This is also expected, and is particularly effective when the condition of the registered image is standard.

In the above embodiment, the case where multiple sets of feature information extracted from finger images for one finger are registered in advance is explained as an example, but the invention is not limited to this. The obtained feature information may also be used. In this case, if the person is not recognized during verification, by replacing the finger with another finger, for example, even if the person cannot be identified with the index finger, the person can be identified with the middle finger, thereby eliminating the person. The occurrence of errors can be reduced.

Furthermore, instead of simply registering a plurality of sets of feature information, only those that have a low correlation with each other may be selected and registered.

Below, a process for selecting and registering n pieces of feature information from g pieces of feature information will be described.

FIG. 5 shows a configuration when applied to a room entry/exit management device, and feature information selection is performed in a feature information selection unit 33 connected to the control unit 25.

FIG. 6 shows the flow of processing related to registration, in which g preliminary registrations are followed by n (n<iJ) main registrations.

First, k as an initial value indicating the number of preliminary registrations is set to 1 (step 5T31).

Thereafter, the input of the finger image IM (x, y) and the extraction of the feature information P (x) are repeated while repositioning the finger F until the initial value k or the predetermined number g is reached (step 5T32).
~36). As a result, g pieces of feature information P□h(X) are obtained.

Pok (X) (x −0,1,2,...,511,k −1,2,
3,...,11) ...(8) The nine pieces of feature information P obtained. , (X) are preliminarily registered in the memory 32 as candidates for feature information to be matched.

When this preliminary registration is completed, the initial value k is reset to 1 again (step 5T37), and then characteristic information for main registration is selected (steps 5738.39, 40.4).
1).

Here, the selection of feature information is, for example, 9 pre-registered
It is determined by combining two IC feature information two by two to create two combinations of ICs, and obtaining a correlation for each. The correlation can be obtained in the same manner as the calculation of the squared error f (t) at a certain position t and the matching value S during matching described above.

That is, f (t) ... (9) S-min (f (t)1 (-tl<t<tl) ... (10)
St , + −V (Poh (X) +
Pot (X) )(k, j-1,2,3,
..., J7, k≠j) ... (11) Furthermore,
An average Qk of correlations between a certain piece of feature information Pot(x) and all other feature information is determined, and the correlation of feature information rom(x) with other feature information is defined by this average Qk.

That is, the average Q is repeatedly determined until an initial value k or a predetermined number g is reached, and n pieces of feature information P are determined in order of increasing value (low correlation) of this average Qk. k(x) is selected.

n pieces of feature information P selected in this way. ,,(x) is registered in the IC card 28 as feature information to be verified Po:(X) (step 5T42).

According to such a configuration, highly correlated features, that is, multiple pieces of feature information that are very similar, are not included, so efficient registration can be achieved without unnecessarily increasing the number of feature information to be matched. 1 during verification.
It is possible to increase the possibility that feature information matching the two pieces of feature information to be compared is included, and it is possible to more effectively reduce errors in determining that the person is a different person.

Furthermore, at least one of the feature information to be matched may be obtained from another hand.

FIG. 7 shows an example of processing related to verification when another method is used for registration.

For example, assume that a finger image IM (x, y) of a certain finger F is input (step 5T51). Then, feature information P(x) is extracted from the finger image IM(x, y) (step 5T52).

This feature information P (x) is collated after positioning with the first feature information to be verified out of a plurality of feature information to be verified registered in advance in the IC card 28 (step 5T53). (Step 5T54).

As a result of this verification, if the person to be authenticated is determined to be the authenticated person (step 5T55), the process is terminated.

On the other hand, if the person to be authenticated is not determined to be the authenticated person, a retry opportunity is given (step 5T56). That is, the person to be authenticated has a finger image IM(x.

After changing the finger F for inputting y) to another hand, the processes of steps 5T51 to 55 described above are repeated. In this case, the first feature information to be matched is discarded and compared with the newly read second feature information to be matched, that is, the feature information to be matched that has been extracted from the finger of the changed hand and registered in advance. be done.

In this way, if feature information extracted from each finger of another hand is registered in advance, even if the hand used for registration is injured, the fingers of both hands can be injured. Unless otherwise specified, it is possible to have at least one valid feature information to be verified, and it is possible to accurately confirm whether or not the person is the person.

Furthermore, characteristic information used for personal authentication includes at least a fingerprint and a first... The present invention is not limited to a one-dimensional signal obtained by adding pixel densities of a finger image including the second joint portion in a direction perpendicular to the longitudinal direction of the finger, but can be similarly implemented using fingerprint information.

It goes without saying that various other modifications can be made without departing from the gist of the invention.

[Effects of the Invention] As detailed above, according to the present invention, it is possible to absorb slight fluctuations in characteristic information due to differences in finger placement, thereby improving verification accuracy and authentication reliability. It is possible to provide a personal authentication device that can improve the following and reduce the occurrence of errors that exclude the person in question.

[Brief explanation of the drawing]

Fig. 1 is a block diagram of an entry/exit control device showing an embodiment of the present invention, Fig. 2 is a diagram schematically showing the relationship between a finger image and feature information, and Fig. 3 is an operation related to registering feature information. 4 is a flowchart shown to explain the operation related to matching, and FIG. 5 shows an example of selecting and registering features with low correlation to each other as feature information to be matched. A block diagram of the entry/exit control device, Fig. 6 is a flowchart shown to explain the operation when selecting and registering characteristic information with low correlation, and Fig. 7 shows another method used for registering characteristic information. FIG. 8 is a flowchart shown to explain the operation related to the case verification, and FIG. 8 is a block diagram of the room entry/exit control device shown to explain the prior art and its problems. 31...Door drive unit, 32...Memory, 33...Characteristic information selection unit.

Claims (4)

[Claims] (1) An input means for inputting image information of a finger including at least the first and second joints; a feature extraction means for extracting feature information from the image information of the finger inputted by the input means; and the input means a feature registration means for registering a plurality of pieces of feature information obtained by repeating an input operation by the user and an extraction operation by the feature extractor as feature information to be matched; a comparison means for continuously comparing the feature information and the feature information extracted by the feature extraction means from the image information of the finger separately input by the input means; and based on the comparison result by the comparison means. A personal authentication device comprising: a determining means for determining whether the characteristic information and the characteristic information to be compared are extracted from the same finger. (2) an input means for inputting image information of a finger including at least the first and second joints; a feature extraction means for extracting feature information from the image information of the finger inputted by the input means; and the input means a feature registration means for registering each feature information obtained from a plurality of fingers by repeating an input operation by the user and an extraction operation by the feature extraction means as feature information to be compared; a comparison means for continuously comparing feature information to be matched and feature information extracted by the feature extraction means from finger image information separately input by the input means; and a comparison by the comparison means. and a determining means for determining whether or not there is any match with the characteristic information in the characteristic information to be matched based on the result, and if the determining means does not determine a match, the comparing means A personal authentication device characterized in that a comparison operation is performed using another finger. (3) an input means for inputting image information of a finger including at least the first and second joints; a feature extraction means for extracting feature information from the image information of the finger inputted by the input means; and the input means a feature registration means for selecting feature information having low correlation with each other from a plurality of feature information obtained by repeating an input operation by the user and an extraction operation by the feature extractor and registering the selected feature information as feature information to be matched; a comparing means for continuously comparing a plurality of pre-registered pieces of feature information to be matched and feature information extracted by the feature extracting means from finger image information input separately from the input means; , a determination means for determining whether or not the feature information and the feature information to be matched are extracted from the same finger based on the comparison result by the comparison means. Authentication device. (4) an input means for inputting image information of a finger including at least the first and second joints; a feature extraction means for extracting feature information from the image information of the finger inputted by the input means; and the input means a feature registration means for registering each feature information obtained from a plurality of fingers by repeating an input operation by the user and an extraction operation by the feature extraction means as feature information to be compared; Comparing means for comparing feature information to be matched and feature information extracted by the feature extracting means from finger image information separately input by the input means, and comparing by the comparing means. A personal authentication device that authenticates an individual based on a result, characterized in that at least one of the plurality of feature information for verification registered in the feature registration means is obtained from a finger of another hand. personal authentication device.