JPH02148182A - Fingerprint collating device - Google Patents

Abstract

PURPOSE:To always accurately perform fingerprint discriminating processes by effectively utilizing a fuzzy theory upon collating a fingeprint image to be collated with each register partial image. CONSTITUTION:A weighting factor deciding means 6 discriminates the ambiguity of degrees of collation in accordance with each number of discordant image element based on the four fuzzy data of a storing means 1 and weighting factors are respectively decided from the pattern data stored in the storing means 1 in accordance with each prescribed value by using the degrees of collation as four prescribed values. An adding means 7 adds the four prescribed values to each other after respectively weighting the values with the decided weighting factors and a discriminating means 5 discriminates the collated state of a binarized image to be collated with fingerprint image information in accordance with the added results of the means 7. Therefore, even when the image quality of the binarized image to be collated changes from that of fingerprint image information at the time of collation, high collating accuracy can be maintained without increasing the erroneous recognition rate.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to a fingerprint verification device suitable for use in an entry control system for an important secret room such as a computer room, or a keyless entry system for a vehicle, and in particular, The present invention relates to a fingerprint verification device that compares a fingerprint image optically input for verification with a registered fingerprint image.

(Prior Art) Conventionally, in this type of fingerprint verification device,
As shown in Publication No. 78286, partial images of a plurality of small areas are cut out from the fingerprint image input at the time of registration and each is used as a registered partial image, and at the time of verification, each registration in the verification fingerprint image input at the time of verification is performed. There is a method in which a partial image is scanned, the degrees of mismatch at the most matching positions are determined, and a final match determination is made by comparing these degrees of mismatch with a single threshold value.

(Problem to be Solved by the Invention) However, in such a configuration, as described above, each degree of mismatch between all the registered partial images and the verification fingerprint image is compared based on only a single threshold value. If the image quality of the fingerprint image during verification changes compared to the fingerprint image during registration,
There is a problem in that the final verification determination result, that is, the verification recognition rate is extremely reduced.

Therefore, in order to deal with such problems, the present invention attempts to effectively utilize fuzzy theory in matching each registered partial image of a verification fingerprint image as described above in a fingerprint verification device. It is.

(Means for solving the problem) In solving the problem, the structural features of the present invention are as follows:
As illustrated in FIG. 1, there is a storage means 1 for storing a plurality of partial images of a binary fingerprint image representing the fingerprint of a specific person as fingerprint image information, and a storage means 1 for optically reading the fingerprint of a person requesting verification and comparing binary values. a reading means 2 for converting the verification binary image into a converted image; a matching processing means 3 for performing a matching process by comparing the verification binary image with the fingerprint image information; and each of the partial images of the verification binary image after the matching process. A fingerprint verification device comprising: a measuring means 4 for measuring the number of mismatched pixels between the two; and a determining means 5 for determining the state of comparison of the verification binary image with the fingerprint image information according to the number of mismatched pixels. , a first pattern in which the two-dimensional relationship between the weighting coefficient and the number of mismatched pixels is defined by a curve based on a predetermined number of mismatched pixels, and a first pattern in which the two-dimensional relationship is defined by a separate curve based on the predetermined number of mismatched pixels. pattern data consisting of the second and third patterns, the number of mismatched pixels and the first and second patterns;
and a third pattern to determine the ambiguity of the degree of matching, and determine the degree of matching by determining a first predetermined value, a second predetermined value with the same sign smaller than the first predetermined value, and a second predetermined value. A third predetermined value that differs only in sign, or a fourth predetermined value that differs only in sign from the first predetermined value, is stored in the storage means 1, and the Weighting coefficient determining means 6 that determines weighting coefficients from the pattern data according to each number of mismatched pixels, and addition means 7 that weights and adds the first to fourth predetermined values with each of the determined weighting coefficients. The present invention is such that the determining means 5 determines the collation state according to the addition result of the adding means 7.

(Function) By configuring the present invention in this way, when the reading means 2 optically reads the fingerprint of the person requesting verification as a verification binary image, the matching processing means 3 stores the verification binary image. Matching processing is performed by comparing with the fingerprint image information of means 1,
The measuring means 4 measures the number of mismatched pixels between the matching binary image after the matching process and each partial image of the fingerprint image information, and the weighting coefficient determining means 6 measures the number of pixels of the matching binary image after the matching process,
Based on the fuzzy data, the ambiguity of the matching degree is determined according to the number of mismatched pixels, and the matching degrees are set as the first to fourth predetermined values, and according to each of these predetermined values, the ambiguity of the matching degree is determined. Weighting coefficients are respectively determined, and the adding means 7 weights and adds the first to fourth predetermined values using the determined weighting coefficients, and the determining means 5 performs the collation and binarization according to the addition results. A comparison state of the image with the fingerprint image information is determined.

(Effect) Therefore, the judgment of the ambiguity of the matching degree as described above,
By determining the first to fourth predetermined values, determining each room filling coefficient, and determining the verification state by the determining means 5 based on the above-mentioned addition results by the adding means 7, verification at the time of verification is performed. Even if the image quality of the binarized image changes from that of the fingerprint image information, high matching accuracy can be maintained without increasing the false recognition rate.

(Example) Hereinafter, one example of the present invention will be described with reference to the drawings.
The figure shows an example in which the fingerprint verification device according to the present invention is applied to a door lock mechanism 10 provided on the door of an important secret room such as a computer room that requires entry control. This fingerprint verification device includes a verification button switch 20, which, when operated, generates a verification request signal. The numeric keypad 30 is operated to generate a designation code signal when the person requesting entry designates his or her own registered fingerprint image information from among a plurality of registered fingerprint image information registered in the memory 50, as will be described later. In such a case, the designation code that is the content of the designation code signal corresponds to the person's own fingerprint known only to the person requesting entry.

The image input device 40 optically reads the fingerprint of the person requesting admission when the finger of the person requesting admission to the room is pressed against the reading screen of the image sensor, and generates the read fingerprint image as an analog signal. In the memory 50, registered fingerprint image information for identifying the fingerprints of a plurality of persons who are permitted to enter the important secret room is stored and registered in advance. In such a case, each of the registered fingerprint image information mentioned above is composed of three registered partial images, and each of these three registered partial images is the fingerprint image of each of the plurality of persons permitted to enter the important secret room mentioned above. This corresponds to partial images of each of three small areas cut out from the binarized fingerprint image respectively corresponding to .

The microcomputer 60 executes the computer program in cooperation with the verification button switch 20, the numeric keypad 30, the image input device 40, and the memory 50 according to the flowcharts shown in FIGS. 3 and 4, and during this execution, Drive circuit 70 connected to door lock mechanism 10
Performs calculation processing necessary for control. However, the above computer program is stored in the ROM of the microcomputer 60.
is stored in advance.

In this embodiment configured as described above, when the device of the present invention is put into operation, the microcomputer 60 starts executing a computer program in step 100a according to the flowchart of FIG. The determination is "NO" based on the non-operation. At such a stage, if the person requesting entry operates the verification button switch 20 and also operates the numeric keypad 30 according to his designated code, the verification button switch 20
A verification request signal is generated from the numeric keypad 30, and the specified code of the person requesting room entry is generated as a specified code signal. Furthermore, when the person requesting room entry presses his or her finger on the reading screen of the image sensor of the image input device 40, the image input device 40 optically reads the fingerprint of the person requesting the coterie room and converts this read fingerprint image into an analog signal. Occur.

Accordingly, the microcomputer 60 performs step 110 based on the verification request signal from the verification button switch 20.
rYES is determined, and in step 120, one registered fingerprint image information is designated and read out of the plurality of registered fingerprint image information in the memory 50 based on the designated code signal from the numeric keypad 30, and in step 130, Image input device 4
0, and step 14
0, the read fingerprint image, which is the content of the analog signal, is binarized to form a binarized fingerprint image.

After that, the microcomputer 60 performs step 150.
At step 140, a matching process is performed on the binarized fingerprint image while moving the three registered partial images of the designated registered fingerprint image information at step 120 so as to minimize the degree of mismatch. When the matching process is completed in this way, the microcomputer 60 measures the number of mismatched pixels 5IS2 and S3 between the three registered partial images after the matching process and the binary fingerprint image in step 170. Fuzzy data R
The specification data i necessary for specifying the fuzzy rule Ri in the middle is set to "1", and the combinatorial program proceeds to the weighting coefficient calculation routine 180.

In this example, the above fuzzy data R is as follows (
13 fuzzy rules Ri (i
=1.2. . . , 13) and are stored in the ROM of the microcomputer 60 in advance.

(Table-1) However, each symbol VS, MD, and VL shown in (Table-1) represents each pattern as shown in FIG. The relationship with the number S of mismatched pixels is specified. Further, the pattern MD is composed of straight lines corresponding to both hypotenuses of an isosceles triangle drawn symmetrically with respect to the straight line S=Smd. In addition, both patterns ■S and VL are
It is composed of straight lines corresponding to the upper base and one leg of each isosceles trapezoid drawn symmetrically with respect to the straight line S=Smd on both sides of the pattern MD.

In such a case, the vertices of pattern MD are (S, Wij)=
(Smd, 1 ), and both lower ends of pattern MD are (S, WIJ) = (Sma, O) and (S, W
IJ) = (Smb, O), and the intersection of the upper base and legs of the pattern VS is (S, WIJ) = (Sma,
1), and the intersection of the upper base and the legs of the pattern VL is specified by (S, WIJ) = (Smb, 1'). Also, 0 where Smd-3ma=Smb-3md
As described above, each pattern V is symmetrical with respect to S=Smd.
The purpose of configuring S, MD, and VL is to improve matching accuracy without increasing the rate of misrecognition in which another person's fingerprint is mistakenly determined to be the person's fingerprint, and requires only high matching accuracy or only a low rate of misrecognition. In such cases, it is effective to set an asymmetric pattern. In addition, pattern S corresponds to the case where the above-mentioned matching processing result is a match, pattern VL corresponds to the case where the matching processing result is a mismatch, and pattern MD corresponds to a case where the matching processing result is a match or a mismatch. This corresponds to a case where neither of the above can be said.

In addition, in (Table 1), the code VR represents the degree of matching of the application results of the fuzzy rule Ri, and each code A, B,
C, D, and E each represent evaluation values for the application results of the fuzzy rule R1.

In such a case, the relationship between the -several VR and each evaluation value A to E is as follows:
Each score Po, P shown in the following table-2).

It is specified by 0, Pl, and Pa.

(Table-2) However, in (Table-2), P o > P 1 > O
shall be.

In addition, each fuzzy rule Ri and (
The relationship between VR and A-E shown in Table 2) is stored in advance in the ROM of the microcomputer 60.

When the computer program proceeds to the weighting coefficient calculation routine 180 as described above, the microcomputer 60 performs calculation processing based on the fuzzy rule R1 in step 181 according to the flowchart of FIG. However,
If the number of mismatched pixels 81 in step 160 is on the pattern vS, the microcomputer 60 determines that 51=VS and VR=A, and based on the relationship in (Table 2), the same VR=PO. Next, in step 181a, the microcomputer 60 determines a weighting coefficient Wl in accordance with the number 81 of mismatched pixels based on the pattern VS. Note that in WI, the code i is the fuzzy rule R
is the same as the suffix i of i, while the sign j is 5=
It is the same as the suffix j of Sj. The weighting coefficient of the fuzzy rule R when S=S is illustrated in FIG. 5 as W=0.8.

Thereafter, the microcomputer 60 sequentially determines each weighting coefficient W12 and Wl in the next steps 182 and 183. In such a case, in fuzzy rule R, S2. Since S, is not defined, w12=w,,
= l, Q is determined. Next, in step 184, the microcomputer 60 determines that W1□≦W12.
and W1! It is determined whether or not ≦W13 holds true. In such a case, if it is W11O08 as mentioned above, W
1□-W ts" 1, 0, so the determination at step 184 is rYESJ. Therefore, the microcomputer 60 determines at step 185 that WI = W + t = 0.8
I decide. However, the code W1 is W, , W, 2 and Wl
, represents the minimum weighting coefficient of . Note that, generally, the symbol W represents the minimum weighting coefficient among wll+W1□ and Wl3.

When the execution of the weighting coefficient calculation routine 180 is completed in step 180b, the microcomputer 6
0, in step 190 of FIG.
W1t=0.8) and VR=P in step 181
Calculate according to o.

However, it is assumed that the preceding addition value VP is initialized in step 100a and vp=o.

VP=VP+Wi
2 and update. Note that equation (1) is stored in advance in the ROM of the microcomputer 60.

As described above, after the addition process in step 210, the microcomputer 60 executes the weighting coefficient calculation routine 180.
is executed based on i=2. In such a case, calculation processing based on fuzzy rule R2 is performed, but fuzzy rule R2
, only S2 is defined, and S, ,S, is not defined.Therefore, in step 181, step 1
If the number of mismatched pixels at 60 is 82 on the pattern VS,
S2-VS is determined as VR=A, and according to this VR=A, V is determined based on the relationship (Table 2).
It is determined that R=P. Also, both steps 181a, 1
In step 83, it is determined that W21==W23=1.0, while in step 182, the weighting coefficient W2□ is determined as pattern ■
Based on S, it is determined according to 5=82. In addition, when the weighting coefficient of the fuzzy rule R2 in the case of 5-82 is illustrated in FIG. 5, W22=0.4.

However, W2□−W2. = 1.0 and W22 = 0.4 as described above, W2□ > W 2□ Therefore, the determination in step 184 is rNo, roar, and W
22(=0.4)<W21(=1.0) and W22=0
．． 4<W23 (=1.0), so microcomputer 6
0 is determined to be rYEsJ at step 186, and determined to be W2-W2□ at step 187. Thereafter, in step 190, the microcomputer 60 calculates the latest addition value VP in the same step and the weighting coefficient W2 = 0 in step 187.

4 and VR=P in step 181 (1)
An additional value vp is calculated based on the formula.

After the addition of i=3 in step 210 following the determination of "NO" in step 200, in the execution of the weighting coefficient calculation routine 180, calculation processing based on the fuzzy rule R3 is performed. In such a case, fuzzy rule R3
, only S, is defined, and S, , S2 is not defined. Therefore, in step 181, step 16
If the number of mismatched pixels 83 at 0 is not on the pattern VS as illustrated in FIG. 5, assuming that VS coincides with the S axis, VR=A=Po is determined as. Moreover, in each step 181a, 182,
After fuzzy rule R3, W,,=W32=1,
On the other hand, in step 183, W, 3-O is determined on the assumption that VS coincides with the S axis.

Therefore, if W33=O as mentioned above in wo,=w,2=1.0, then step 1
The determination at 84 is “NO”, and W3□(-1,
Since 0)>Wl3 (=O), the microcomputer 60 determines rNOJ at step 186, and determines W3=W33=0 at step 188. Thereafter, in step 190, the microcomputer 60 calculates the added value VP based on equation (1) according to the latest added value VP in the same step, W3=0 in step 188, and VR=P0 in step 181. do.

After the addition of i=4 in step 210 in conjunction with the determination of rNOJ in step 200, in the execution of the weighting coefficient calculation routine 180, calculation processing based on the fuzzy rule R4 is performed. However, in step 181,
If each mismatched pixel number S, , S2 in step 160 is on each pattern MD, VS, then S, = MD and S,
=VS, and the number of mismatched pixels s3 at step 160
Assuming that the relationship between and VS is the same as in the case of fuzzy rule R3, VR = A = P
, is determined. Further, in step 181a, the weighting coefficient W41 is determined according to SmS based on the pattern MD, and in step 182, the weighting coefficient W42 is determined according to the pattern vS.
Based on VS=O, the weighting coefficient W43 is determined to be 5=82, and in step 183, the weighting coefficient W43 is determined to be zero based on VS=O.

The weighting coefficients of the fuzzy rule R4 when 5=81 and S2 are illustrated in FIG. 5 as W41=0.2 and W42=0.6.

As mentioned above, W41=0, 2, W42=0,
6 and W43=0, the microcomputer 60 at step 184 sets W4.6 and W43=0. >Because of W43, [
It is determined No, and in step 186, W 42 >
W4□ and W4□〉W4. Therefore, it is determined that it is rNOJ, and in step 188 it is determined that W 4 =W 43 =O.

After that, the microcomputer 60 performs step 190.
, the latest added value VP at the same step, step 18
According to W4=O at step 8 and VR=P0 at step 181, an additional value vp is calculated based on equation (1).

Hereinafter, each fuzzy rule R5 to
Based on R13, the weighting coefficient calculation routine 180 is repeatedly executed in substantially the same manner as described above. In such a case, as shown in (Table 1), VR
=B, VR=C for fuzzy rule R7, VR=D for both fuzzy rules R8゜R9, remaining fuzzy rule R
In IO to R13, VR=E, and as shown in Table 2, VR=B, C, D, and E.

=p, o, -p, -p, roar. Conoco means that VR takes values that are approximately symmetrical in positive and negative values centering on the case of fuzzy rule R7. Also, both patterns VS and VL are symmetrical to SmSmd, and Gatsu pattern MD is also SmSm
Since it is symmetrical to d, Wl is the total 7y G rule Rl
~Rt3, and exist almost symmetrically in each range of S<Smd and S>Smd. Further, each time the weighting coefficient calculation routine 180 as described above is executed, the addition/updating process of ■P in step 190 is repeated in the same manner as described above.

After that, when the determination in step 200 becomes "YES", the microcomputer 60 determines in step 220 whether the latest addition update VP in step 190 is positive or negative. Therefore, if vp>o, the determination at step 220 is rYES. Therefore, based on the judgment that the weighting coefficient WIJ (ie, W+) in the range S<Smd in FIG. 5 has a high influence on vp, the microcomputer 60 in step 230 performs It is determined that the digitized fingerprint image matches the designated registered fingerprint image information in step 120.

Then, the microcomputer 60 performs step 240.
An output signal for releasing the door lock of the door lock mechanism 10 is generated, and in response to this, the drive circuit 70 releases the locked state of the door lock mechanism 10. On the other hand, if the determination in step 220 is rNOJ, the fifth
In the figure, the weighting coefficient W1 in the range S>Smd. (i.e. W
Based on the judgment that I) has a high degree of influence on ■P,
At step 250, the microcomputer 60 determines that there is a mismatch, contrary to the case at step 230, and at step 260, the output signal remains extinguished and the drive circuit 70 is prohibited from unlocking the door lock mechanism 10. .

As explained above, when verifying the fingerprint of the person requesting entry, each pattern MD, VS, VL, the contents of each fuzzy rule R1 to R13 (Table 2), and each weight coefficient calculation routine 180 and step 190 are used. The calculation processing contents are the number of mismatched pixels St, s2 at step 160.
．． Even if there is variation in S3, each step 220, 23
The determination process at 0.250 can be performed appropriately. This means that even if the image quality of the binarized image in step 140 differs from that of the fingerprint image information in the memory 50, the above-described determination process can be performed with high accuracy without increasing the rate of misrecognition. . As a result, the matching recognition rate between the person authorized to enter the room and the person in question increases, while the misrecognition rate between the person and another person decreases. Therefore, the door lock mechanism 10 will not be unlocked by mistake, and each step 184 to
In 188, W,,,W,2. Since the smallest of W and g is set as WI, reliability can be ensured in determining the weighting coefficients.

Further, in implementing the present invention, each partial image constituting the fingerprint image information in the memory 50 may be at least two or more, and accordingly, the number of mismatched pixels,
The contents of the fuzzy rules may also be changed as appropriate.

In addition, in carrying out the present invention, each pattern MD in FIG.
, VS, and VL may be changed as necessary.

Further, in implementing the present invention, the present invention is not limited to important secret rooms, and may be applied to keyless entry systems of vehicles and the like.

[Brief explanation of the drawing]

FIG. 1 is a diagram corresponding to the claims, FIG. 2 is a block diagram showing an embodiment of the present invention, FIGS. 3 and 4 are flow charts showing the operation of the microcomputer shown in FIG. 2, and FIG. The figure is a characteristic diagram showing the relationship between the number of mismatched pixels and the weighting coefficient. Fig. 1 40. 0. Explanation of symbols: Image input device, 50. Microcomputer.・Memory, 6

Claims (1)

[Claims] a storage means for storing a plurality of partial images of a binarized fingerprint image representing the fingerprint of a specific person as fingerprint image information; a reading means for optically reading the fingerprint of a person requesting verification and converting it into a binary image for verification; and said verification means. A matching processing means that performs matching processing by comparing the binary image with the fingerprint image information, and a measuring means that measures the number of mismatched pixels between each of the partial images of the matching binary image after the matching processing. and a determining means for determining a matching state of the matching binary image with the fingerprint image information according to the number of mismatching pixels, wherein a two-dimensional relationship between a weighting coefficient and the number of mismatching pixels is provided. is determined by a curve based on a predetermined number of mismatched pixels.
pattern, and pattern data consisting of second and third patterns in which the two-dimensional relationship is defined by separate curves based on the predetermined number of mismatched pixels, and each number of mismatched pixels and the first, second, and third patterns. The ambiguity of the degree of matching is determined using the pattern of , and the degree of matching is determined using a first predetermined value, a second predetermined value of the same sign smaller than this first predetermined value, and this second predetermined value.
A third predetermined value that differs only in sign from the predetermined value, or a fourth predetermined value that differs only in sign from the first predetermined value is stored in the storage means. , and weighting coefficient determining means for determining weighting coefficients from the pattern data according to each number of mismatched pixels; and addition means for weighting and adding the first to fourth predetermined values using each of the determined weighting coefficients. A fingerprint verification device characterized in that the determination means determines the verification state according to the addition result of the addition means.