JPS59778A - Finger print collating device - Google Patents

Abstract

PURPOSE:To evaluate couples of characteristic points ''couple'' corresponding to two finger prints, by calling an area pattern assigned to a memory, a characteristic point list, a deviation plane, a candidate couple list and a list with a processor for processing them for the purpose of collation. CONSTITUTION:Each absolute value is calculated with absolute devices 154x, y, d, z to two characteristic data X, Y inputted via signals 151x, y, d, z and 151x', y', d', z' from a control section. Further, a threshold value group selected with a signal 150 from the control section is compared with outputs 153x, y, d, z of a threshold value ROM 153 at comparators 155x, y, d, z. The output of the comparison is combined at an AND 156 and whether or not all conditions are satisfied is checked and returned to the control section as an output 152.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a device that verifies the identity of a pattern composed of a striped pattern such as a fingerprint, based on the characteristics of the pattern.

Conventionally, as a feature of verification of striped patterns such as fingerprints, as shown in FIG. 1, the positions (4, Yρ direction Df,
It was proposed to use relations (Mr, , R (r; r = 1 to 4) and density Cf. A device for extracting these features is described in Japanese Patent Application No. 54-39648 (Japanese Unexamined Patent Publication No. 56-198).
-138174 Publication), and Japanese Patent Application No. 54-98966 (Japanese Unexamined Patent Publication No. 56-2467) regarding the verification device.
No. 5). Here, the relation refers to the local coordinate system (ξ,
The number of ridges between the nearest point Mfr and the origin Mf in each quadrant of This is a quantification of the flute quality.However, the above-mentioned patent application
The verification device disclosed in the specification of No. 4-98966 does not strictly examine the correspondence of individual minutiae in the two fingerprints to be verified, so the identification ability deteriorates when the number of horizontal files becomes large/large. The big problem was that it was not sufficient.

An object of the present invention is to provide a device that enables more efficient fingerprint matching by strictly evaluating the corresponding special points [pairs] in two fingerprints to be matched. .

Next, the present invention will be described and explained in detail with reference to the drawings.

Before explaining the device, we will define the data format of the "E pair of fingerprints to be compared." The characteristics of the fingerprint to be searched are the area pattern PN and the minutiae list I-(Ns; s=1 to S)9, and the characteristics of the file fingerprint to be searched are the area pattern P.
M and feature point list) (f = 1 to F in M). The formats of both features are similar, and only -C of the file fingerprint is shown in FIGS. 2(a) and 2(b).

The contents of (b) (feature point list in Figure 2) are explained in 1.
This has already been explained in connection with the figure. On the other hand, the area pattern shown in FIG. 1(a) is a binary pattern indicating the pattern stamped area shown in FIG. -758
This can be realized by the apparatus shown in Japanese Patent Publication No. 24). Here, it should be noted that feature points are extracted only in the clear region indicated by "1" in the region pattern, and are not extracted in the unclear region indicated by 60".

FIG. 3 is a block diagram showing the configuration of an embodiment of the fingerprint matching device according to the present invention. For the control unit 1o, an area pattern IIN for storing the characteristics of the search fingerprint and a feature point list) 1
2N, area pattern II for storing file fingerprint characteristics
M and minutiae list) 12M, paired inspection circuit 15 opposed testing circuit 16, deviation plane 17 that stores intermediate result information, paired candidate list 18, paired squirrel that holds paired minutiae of search fingerprints and file fingerprints) 19N, 19M are connected by each input/output data signal line group.

Further, as interface signals from the entire verification system, an input signal 100 for inputting characteristic data of search fingerprints and file fingerprints and an output signal 101 for outputting verification results are provided. As shown in Figure 2(b), X, Y, D,
C, M, , R, (r = 1 to 4) information is retained, among which data X, which depends on the coordinate system expressing the feature points,
Coordinate conversion circuits 13N and 13M are provided for Y and D.

FIG. 4 is a flowchart illustrating an overview of the operation of the collation device shown in FIG. 3. First, when the area pattern and minutiae list, which are the characteristics of the search fingerprint and file fingerprint to be collated, are input into this matching device, all minutiae registered in each minutiae list are brute-forced by their relations. Inspect and store potentially corresponding feature point pairs in a candidate pair list. Next, the amount of coordinate matching where the corresponding feature points most closely match is determined according to this candidate pair, and the coordinates of the feature point list of the file fingerprint are matched according to this amount of matching. Next, the feature point list after coordinate matching is inspected in terms of its arrangement, and candidate pairs are carefully selected. Furthermore, based on the carefully selected candidate pairs, the correspondence by relation is checked, and the candidate pairs given to each candidate pair are corrected. After generating a pair list that determines the final corresponding feature points from the candidate pair list obtained in this way, the pair values are corrected again by the relational controller, and the area pattern and test density of opposing feature points are inspected. to determine the final value of the pair. The pair values are totaled and a matching value normalized by the number of matching feature points is calculated and output as the final output of the matching device. Note that if an appropriate matching amount is not found during coordinate matching, the matching process is omitted and the mismatch matching value is output as the matching value, thereby improving the matching speed.

Hereinafter, the configuration and operation of each part will be explained in detail in the order of the above operation outline.

Referring to FIG. 3, when the area pattern PN of the search fingerprint to be matched and the minutiae list (Nil; s=1 to S) are inputted via the input signal 100, the control unit changes the area pattern IIN to Address ll0N and write signal 111
A region pattern PN is written and held via N. Also, the feature point list is also the feature point list) Address 12ON to 12N
and write and hold via the write signal 121N.

Next, the area pattern PM of the file fingerprint and the minutiae list (M,; f=1 to F) are similarly changed to the area pattern IIM,
Write and hold the feature point list 12M via addresses 110M, 120M and write signals 111M, 121M. Area pattern 11M, IIN is shown in Figure 2 (i)
This is a normal memory that holds binary data in a two-dimensional lattice shape, and the addresses ll0M and ll0N selectively address lattice elements. Also feature point squirrel) 12M
, 12N hold the feature point data shown in FIG. 2(b), and the address 120M, 12ON is the feature point M.
,,N, is a normal memory specified as a unit. These memories are already known and do not require any special explanation. Once the search fingerprint and file fingerprint to be matched are stored in this way, the actual matching operation begins, and the first step is to generate a candidate pair list.

That is, for each minutiae point N, (s=1 to S) of the search fingerprint, the control unit 10 brute-forces each minutiae point M, (f=1 to F) of the file fingerprint via the address 12ON, 120M. Read out.

The own position (x, y) and direction of the feature point data are sent to signals 132N and 132M via coordinate conversion circuits 13N and 13M.
The dense amount C relation (Mr e Rr
; r = 1 to 4) are read out from the signals 122N and 122M, but in this case, the amount of coordinate transformation is determined in advance by the signal 13O.
(ξ=0.η-00-〇) is set via N and M.

FIG. 5 is a block diagram showing one embodiment of the coordinate conversion circuit 13N. Note that 13N and 13M have the same structure. In FIG. 5, the coordinate conversion circuit includes registers with selection inputs 134x, y, d, subtracters 135x, y, d, sine and cosine constant ROM 133, multipliers 136c, s, 137c, s.
and adder 138x.

It is composed of a subtracter 138y. Coordinate transformation amount (ξ, η,
θ) is the signal 130 (130M or 13
ON) or as feature point data X, Y, D signals 131x, y.

The register 134
, Ypd. Also, during the conversion operation, when the feature point data
, y, d (X-ξ), (Y-η), (D-〇)
are output, respectively, and the sine cosine ROM 133 addressed by (D-0) of the signal 132d stores in advance cosine value cos (1)-0) and sine value sin (1
)-〇) are output as signals 1330c and s. Then, multipliers 136c, s, 137C- and adder/subtractor 138x,
By y, ma = (xゝξ)c, oo(D-θ) + (Y-η) sin
(D-θ)mer(Y-η)coo(D-θ)-(X-ξ
) sin(1)-θ) is calculated and output as the output signal 132yrY. The above outputs 132x, y, d are shown in FIG.
It corresponds to N and M. This concludes the explanation of the coordinate conversion circuit.

Now, since the coordinate transformation amount (ξ=η=θ=0), X=X
,, Y=Y, T)=D, meaning no conversion. The feature points Ns, M (two sets of X, Y, D, and C) read out in this way are immediately sent to the paired inspection circuit 15 by the signal 15.
1.

FIG. 6 is a block diagram showing one embodiment of the pair test circuit 15. The test circuit has a threshold R that holds several threshold groups.
OM153 absolute value unit 154XIYIZ, comparator 155x
, y, d, z and AND156.

The operation is performed by signals 151x, y, d, z and 151. x'.

y′,dl,! Absolute value units 154 x, y, d, z calculate the respective absolute values for the two feature point data X, Y, and DBC inputted via (corresponding to 151 in FIG. 1), and the results are sent to the control unit 10. The dark value group selected by the signal 150 is compared with the output 153x, y, d, z of the threshold value ROM 153 by the comparator 155x, y, d, z. This is IX[-X, 1≦Tx, IYf-Y81≦T, ,
IDf-D51fd, meaning lCf-Cs1:5ITc. The output of this comparison result is integrated by AND 156, and a test output 152 determines whether all the above conditions are satisfied.
The data is returned to the control unit 10 as follows. This concludes the explanation of the test circuit.

Minutiae data X, Y, D of search fingerprint and file fingerprint.

When C is tested using the pair test circuit described above, only Tc is strict, and since Tx, Ty, and Td are before coordinate matching, the capacitance condition is such that the fingerprints to be compared are not upside down. Should be set. If the inspection is successful in this way, the control unit 10 controls each coordinate conversion circuit 13N,
The conversion amount registers of M (134x, y, d in FIG. 5) hold the X, Y, and D currently read from the feature point list, and the relation (NSr
=1~4) CM, tr-Rfr; r=1~4)
is held inside the control unit. Then, address 12ON, 1 for minutiae list) 12N, 12M for each quadrant r.
N5r9Mfr is outputted via signal 20M, and its feature point data X, Y, and D are read out via signal 132N and 13th order. In this case, the read feature point data is transformed into the local coordinate systems of the feature points Ns and Mr by coordinate conversion circuits 13N and 13M, respectively.
It has been converted with . This converted feature point data X, Y
, D and the ration R held in advance in the control unit.
are sent to the pair checking circuit 15, and the difference therebetween is checked.

Namely, Shinkan 茫..., Daibi Rika 9., Daiko〈Touch, 6.h.-8.
This means that Iゞ1・' (TaMl, l''f) means the feature imaginary, local coordinate transformation of %lf) has been inspected.The control unit 10 performs this operation for each quadrant r=1 to 4. Do it sequentially,
The number of successes is taken as candidate vs. France W, and if the value is greater than or equal to a predetermined number, candidate vs. feature point Mf and candidate vs. France W for feature point Ns are output to the candidate pair list 18.

To aid understanding, for Ns:Mf as shown in FIG. 7, r=3, that is, NS3:M (only 3 is negated by the test, and in this case the candidate value is W=3.

The candidate pair list 18 may be a normal memory for holding contents as shown in FIG. 8, and as shown in FIG. A set of M5 via signals 181 and 182
i and Wi can be written and read.

When storing the candidate pair feature points Ml, W, the control unit 1
0 performs read/write operations on the above NS line and writes the WS
1≧W, ≧・・・≧”si −”Wsi++≧・・・・
Change the line contents so that W5. Sort and store in order of size.

As described above, a candidate pair list is sequentially created through NSXMF operations, and when this list is completed, a coordinate system matching process using the deviation plane 17 continues.

The deviation plane is located at address 170 as shown in FIG.
It is a normal memory in which writing and reading are performed using the two-dimensional addresses ΔX and ΔY quantized by , and a part of it can also be used with the quantized one-dimensional address ΔD as shown in FIG.

The control unit 10 sends N, (8=1-N, ) and 1 (i=1.2...) to the candidate pair memory 18 via an address 180.
) are sequentially generated and their contents Msi Wgi are read, NS is supplied to the minutiae list) 12N, Msi is supplied to the minutiae list) 12M via addresses j2ON and 120M, and the coordinate conversion circuits 13N and 13M Assuming no transformation, the direction of the feature point ■) S + DS! Read out, △D=D, 1-
Add Wsrt to the element of the deviation plane corresponding to DS.

By repeating this operation, deviation planes for all NS:M5i candidate pairs stored in the candidate pair list become as shown in FIG. 9. In this figure, the black dots schematically show the magnitudes of W and p, and in reality, %i is accumulated in the quantization unit.

If the search fingerprint and the file fingerprint are the same fingerprint, the candidate pair Ns: Msi registered in the candidate pair list has a high probability of being a true pair according to the lation test, and the minutia direction of such a candidate pair is Ds+Dsi is a constant coordinate deviation, r&
It is expected that they will be concentrated as shown in Figure 9. On the other hand, it is of course possible that false pairs exist in the candidate pair list, but these deviations are statistically scattered on the ΔD deviation plane and are not concentrated. The control unit 10 reads out and compares each element of the deviation plane 17, sets the maximum concentration point as the deviation ΔD, and immediately sends a signal 130M only to the coordinate conversion circuit 13M for ξ=O2η−〇, θ=ΔI). Set via.

This means that the search fingerprint and file fingerprint match only with respect to coordinate rotation.

The control unit 10 generates concentration points of the deviations ΔX and ΔY on the deviation plane 17 as shown in FIG. 10 by the same process as the deviation ΔD, and scans and determines the concentration points ΔX and Δy%. , coordinate matching amount ξ−ΔX”, η=ΔY, θ=Δ
1) The coordinates are finally set in the coordinate conversion circuit 13M. Coordinate alignment is completed in the above manner, but if the deviation △ and Δχ and ΔY are determined, if the cumulative density of the concentrated part of the deviation plane is smaller than the preset threshold, the search It is determined that the fingerprint and the file fingerprint are not the same fingerprint, and a negative constant comparison value is output to the output signal 101 to perform the comparison and abort. Also, when enough eggs are obtained, the above coordinate matching amount (ξ, η
, θ), the coordinates of X, Y, and D of the feature point list 12M are transformed, thereby aligning the coordinates of the feature point list. This operation can be carried out by simply reading the address 120M and Mf (f=1 to 4) by the control unit 10 once via the signals 132M and 122M, and then writing again via the signal 121M.

This concludes the explanation of the coordinate matching process.

Next, the control unit 10 resets the coordinate conversion circuit 13M to the non-conversion state, and then converts all Ns in the candidate pair list 18:
Msl is read out again, and its feature point data X, Y, D are tested using the pair testing circuit 15 using thresholds Tx, T, , Td that are stricter than when generating the candidate pair list. This is because, since the search fingerprint and the file fingerprint have already completed coordinate matching, the minutiae should already have the same configuration except for the distortion of the fingerprint imprint. The candidate pair Ns, Msi that is denied in this pair test is deleted from the candidate pair list along with its candidate pair France Wsi. This is the selection process for the complement to 1F'.

Furthermore, in the carefully selected candidate pair list, all candidate pairs N8:M7 are modified as the following candidate pairs with France. That is, referring to FIG. 11, one candidate pair N, :Mf
For each hN, , M(at+zx12ON.

120M, relation (Nsr, Mfr
: r = 1 to 4) and Nsr' of the candidate pair list 18.
Check whether Mrrlj: is registered in IT.

If it is registered as Mf'-MfI, the candidate France Wb is used as the base N5. Add to Ml's candidate vs. France Wa. When the qualification of the candidate pairs is completed, the candidate pair list is sorted every N3 rows according to the size of the new candidate pairs.

With the above steps, the final candidate pair list is completed. Based on this candidate pair list, pairs (squirrels) 19N and 19M are generated.

19N and 19M have similar structures, and as shown in FIG.
M is specified, the feature point number and the French Mf, vs and NS,
There are two sets of memories that can hold Mf.

The generation of 19N and 19M (for squirrels) is initially initialized with negative constants for French squirrels, and then the candidate pair list N, (s-1
-S) multiple times in the order of Ns: M5. That candidate vs. France WS Tomin and the second candidate vs. France W32.1! : Depending on the difference, from the most reliable to the squirrel) 19N, 1
Move that candidate vs. France to 9M as vs. France. The candidate pairs registered in the pair list are deleted from the list of all candidate pairs.

When the contents of all candidate pair lists are moved to the pair list, the pairs of pairs N, :Mf are υ, = υr = (Ns”Mt )
The candidates are against France. Further, the negative initial value of the feature points in the pair list that are not transferred by the candidate pair remains as is.

Next, the decision on whether to deal with France is made by two processes: modifying what is against France and mitigating what is not against France.

(vs. squirrel) Check 19N and 19M sequentially, and if the vs. vs. υ is positive, modify the vs. vs. That is, as shown in FIG. 13, when the vs. French υ5 of N5 in 19N (for example vs. squirrel) is positive, the paired minutiae M (based on the candidate minutiae point squirrel) is Access 12N and 12M and check their relations (Nsr, Msr "r" 1~
4), and check whether the pair N5r in the pair list is M, rtl' and the pair υsr is positive. If true, against France υ5. is added to υ5 against France. Similar processing is performed on the squirrel (19M) side and independently of the squirrel (19N) side.

On the other hand, when a positive pair is not stored in the pair list as shown in FIG.
20M, and the feature point data of Mf' is supplied to the signal 132M.
, 122M and its (XrZ Yt'
) as the two-dimensional address of the area pattern PN and the signal 11.
Supply ON. As a result, if the area value read from the output 112N is "0", it means that the minutiae point Mf' is within the unknown region on the search fingerprint side.
``t care'' is rewritten to a value ``OI''.

On the other hand, if the area value is "1", the value υf' for France remains as is.

Conversely, when performing the above processing with 19N, the coordinate transformation circuit 13N has an inverse transformation amount (ξ--Δχ).

η--ΔY, θ--ΔD*) are set, and the area pattern 11M is examined.

Feature point M where υf′ or υ5′ is kept as a negative value
Regarding f' or N5', the following test is performed. For example, for Mf', its relation (M'(,:
; r = 1 to 4), the counter inspection circuit 16 is used to search for those with Rf+' = 0, which are located close to each other and whose directions are opposite to each other. If there is, the feature point Mfr is checked in 19M, and if the feature point υff′ is a negative value, this feature point Mfr and the above M(
Don't Caye for both 'CD vs. Value vs. France' and change it to "0" for France. As shown in Fig. 15, this is a test of opposing adjacent minutiae points, and these two sets of minutiae may or may not be extracted depending on the condition of fingerprint imprinting or mechanical automatic minutiae extraction. It is an unstable feature point.

FIG. 16 is a block diagram showing one embodiment of the counter inspection circuit 16. That is, absolute difference value unit 163x, y, d, complement value 163c comparator 164x mountain d, direction L[1M 165 absolute difference value unit 166a, b, comparator 167a, b, AN
It is composed of a D gate 168 and a threshold ROM 169.

1 Opposite] Two sets of minutiae data to be inspected (x.

When input signals 161x, ytd and 161x', y', dK are supplied to input signals 161x, ytd and 161x', y', dK respectively, only the direction D' is the direction complement, that is, with the complement value 163C.
Post-difference absolute value unit 163x, y+ inverted by π radians
The absolute value of the difference is calculated at d, and compared with the threshold value output from the threshold ROM 169 by comparators 164x, y, and d. That is, 1x-x'l≦Tx, 1y-y'l≦ry, ID
-1)'+f I≦Td is checked and the output is AND
Gate 168 is manually operated. Here, in the direction component, π radian is connected so as to be opposite to the most significant bit (MSB) of the difference absolute value unit 163d, and correct calculation with respect to periodicity in the direction calculation is maintained. On the other hand, the difference absolute value unit 163x
The difference absolute value output of rY and the difference sign (sign at the time of subtraction) are input as an address in the direction 1165, and the specified direction ΔD is outputted and supplied to the difference absolute value units 166a and 166b. That is, from the fact that △D is close, by inputting all combinations of relatively small x-x' and Y-Y', R
It has been converted to OM. - The absolute value of the difference between the force direction D-D' and the above △D is calculated by difference absolute value units 166a and 166b, and compared with another predetermined value from the threshold ROM 169 by comparators 167a and 167b, The result is provided to AND gate 168. AND gate 168 outputs an "opposite" test affirmation signal when all inputs are asserted at signal 16.
2 to the control unit 10.

In the above-mentioned difference absolute value calculators 166a and 166b, the direction and ΔD
Since the difference of π radians between D' and ΔD corresponds to O radians, the most significant bit (MSB) for subtraction is wired so as to be d.

An example of the opposing inspection circuit has been described above.

In the examination of the pair list with negative pairs, the above-mentioned
If the feature point is not the "opposite" feature point, finally check the density C of that feature point, and if it is very large by value comparison, it will be compared to a value close to "0゛" or negative "θ'". In summary, as schematically shown in Fig. 17, the non-French relaxation exists in the non-common area of the area pattern PN * PM after coordinate matching between the search fingerprint and the file fingerprint. d for non-paired minutiae, “opposed” minutiae indicated by dotted lines in the common area, and minutiae with large density.
It is intended to be relaxed to on'tellre or a negative anti-French close to it.

When the comparison of the pair list to France is completed as described above, the control section 10 calculates XF as the matching value q while sequentially reading the pair list, and outputs it via the output signal 101. Here, 8 and F indicate the number of feature points in the common area of the search fingerprint and the file fingerprint. These operations can be performed using general arithmetic operation circuits, so the details will be omitted.

The detailed explanation of one embodiment of the fingerprint matching device according to the present invention has been completed above, and the configuration of the control unit 0 can be easily accomplished by those skilled in the art based on the explanation of the operation so far. does not require Also, in order to simplify the explanation, the candidate pair list of the candidate pair list and the French modification of the pair list correspond with relay silon (Nsr l Nf
We have only explained the correspondence test for r tr = 1 to 4), but in general, it is not necessarily possible to expect the correspondence of N s, & Mfr at the same r due to the distortion of the configuration of the minutiae points of the fingerprint. Mfr′ with different r′ for
may correspond, so (NsrrMfrtr=1
~4) is not compared in quadrant r units (Nsr;
= 1 to 4) and (Mfr' t r = i to 4), 16
It is accurate to check the correspondence in a round-robin manner. In this case, the correspondence check is carried out using the local coordinate system based on N, , Ml, respectively). These are the coordinate conversion circuit 13N
, 13M, and the pair checking circuit 15, it is possible to perform the same process as when generating the candidate pair list.

As is clear from the description of the operation of the embodiments of the present invention, a commercially available microcomputer is used to control the counter-inspection circuit, counter-inspection circuit, and control section in the processing device.
Furthermore, it is possible to obtain equivalent outputs by allocating region patterns, feature point lists, deviation planes, candidate pair lists, and pair lists using memory, and the present invention does not deal with such changes in hardware configuration. The scope of the claims shall not be limited.

[Brief explanation of the drawing]

FIG. 1 is a diagram for explaining fingerprint features, FIG. 2 is a diagram for explaining a region pattern and feature point list that define fingerprint features, and FIG. 3 is a diagram for explaining an embodiment of a fingerprint matching device according to the present invention. 4 is a schematic flow diagram schematically showing the operation of the present invention. FIG. 5 is a block diagram showing an example of a coordinate conversion circuit. FIG. 6 is a block diagram showing an example of an L pair inspection circuit. , FIG. 7 is a diagram explaining the contents of candidate pair inspection, FIG. 8 is a diagram showing the structure of a candidate pair list, and FIG. 9 is a diagram showing a deviation plane showing direction matching in coordinate matching. , FIG. 10 is a diagram showing a deviation plane showing the deviation matching of the above, FIG. 11 is a diagram explaining the candidate pair list of the candidate pair list, FIG. 12 is a diagram showing the structure of the pair list, Figure 13 is a diagram explaining the French modification of the pair list, Figure 14 is a diagram explaining the non-opposing minutiae of the pair list, and Figure 15 is a diagram explaining the shape of the opposing minutiae on the fingerprint. FIG. 16 is a block diagram showing an example of a one-way inspection circuit, and FIG. 17 is a diagram showing an example of feature points to which anti-French mitigation is applied. In the figure, a control unit 10, a search fingerprint area pattern IIN, a minutiae list 12N, a file fingerprint area pattern 11M, a minutiae list 12M, coordinate conversion circuits 13N and 13M, and a "pair"
Inspection circuit 15, "opposite" inspection circuit 16, deviation plane 17,
Candidate pair list 18 and pair squirrels) 19N and M are shown, respectively. Shi-X -494- N5 and No. 495

Claims (1)

[Claims] In a fingerprint matching device that determines whether two collected fingerprints are the same fingerprint based on the pattern characteristics of the two fingerprints, an area pattern and feature point list storage means that retains the two pattern characteristics. and a paired inspection circuit for inspecting the consistency of minutiae points among the pattern features, and for detecting the amount of coordinate transformation for two-dimensional alignment of the fingerprint using the pair of minutiae points that are candidates by the testing circuit. a deviation plane storage means for storing paired minutiae points as candidates, a candidate pair list storage means for managing the group of paired minutiae points to be candidates, a pair list storage means for selecting and managing true paired minutiae points from the candidate pairs, and It is composed of an opposing inspection circuit that inspects opposing minutiae points, and a control unit that integrally manages the storage means and inspection circuit, and after selecting two candidate pairs of fingerprint features, performs coordinate matching, and then After determining the true pair from , the matching value of the two fingerprints is determined from the pairing value and the number of matching effective minutiae points assigned to each pair and non-pairing minutiae obtained by inspecting the unpaired minutiae. A fingerprint verification device characterized by calculating.