JPH0624002B2 - Fingerprint matching device - Google Patents

Description

TECHNICAL FIELD The present invention relates to a fingerprint collation device.

(Prior Art) Conventionally, for example, in automatic collation of a seal or a fingerprint, a reference point such as a center coordinate of an object to be collated is detected for the purpose of performing the collation accurately and quickly.

For this purpose, various devices for detecting the center of a figure such as a seal stamp have been proposed. For example, Japanese Patent Laid-Open No. 58-191084 filed by the applicant of this application detects the coordinates of the center point of the imprint. A pattern recognition device suitable for use in the above is disclosed.

The apparatus described in the above-mentioned publication reads a figure such as a seal imprint by a raster scanning method and gives a binary electric signal corresponding to a black point and a white point of each coordinate, and a reading section of the figure read by the reading section. It is provided with a fill pattern creation unit for filling the inside, and a center coordinate detection unit for using the barycentric coordinates of each weighted point of the filled pattern as the center coordinates of the figure. According to this device, since the center point of the figure pattern is detected by calculating the center of gravity after painting the figure pattern, the detection error caused by the imprint pattern, the figure pattern blurring, the stain, and the internal pattern unevenness is extremely reduced. Therefore, it is possible to detect the center point coordinates with high accuracy.

(Problems to be Solved by the Invention) However, in the conventional device having the above-described configuration, when the center point is obtained for a figure such as a fingerprint that does not have an outwardly convex closed curve, There is a problem that the point detection accuracy decreases.

More specifically, for example, stripes (line patterns) such as fingerprints
In the case of a part-shaped figure, if there is a break in a part of the outermost line, or if a part of the line is concave inside the figure and a part of the figure is missing In these cases, the position of the center of gravity may be displaced due to these effects, which lowers the accuracy of detecting the reference point such as the center point, and therefore the fingerprints to be collated can be accurately aligned. It will not be possible.

An object of the present invention is to solve the above-mentioned problems and to provide a fingerprint collation device capable of collating fingerprints with high alignment accuracy.

(Means for Solving Problems) In order to achieve this object, the fingerprint collation device of the present invention includes a reading unit for reading a fingerprint to be collated as a binary collation pattern, and a collation pattern memory for storing this collation pattern. , A registered pattern memory that stores a pre-registered fingerprint as the above-mentioned binary registered pattern, and the above-described matching pattern memory and registered pattern memory are scanned in a raster method in each of the first direction and the second direction. The line number distribution extraction unit that obtains the line numbers of these patterns for each scan and extracts the line number distributions of the both patterns in the first and second directions, the collation pattern memory, and the registered pattern memory described above. In the first direction and in the second direction respectively in a raster system, and the maximum line widths of these patterns are obtained for each scan. A line width distribution extraction unit that extracts the line width distribution for each of the first and second directions, and the line width distributions between the above-described matching pattern and the above-described registered pattern and the number of lines for each of the above-mentioned first and second directions. A reference point extraction unit that compares the distributions with each other and extracts the overlapping position of both patterns, and the above-mentioned matching and registered pattern based on the above-mentioned overlapping position, finds the degree of coincidence of both patterns, and obtains this degree of coincidence. It is characterized by comprising a coincidence determination unit for determining whether or not the two patterns match based on each other.

In carrying out the present invention, the above-mentioned reference point extracting unit compares the above-mentioned first-direction line number distributions and line-width distributions of both patterns with each other with the minimum degree of difference and It is preferable to determine that the two patterns do not match when at least one value of the minimum difference obtained by comparing the line number distributions in the second direction and the line width distributions in the second direction is larger than a predetermined value. is there.

In practicing the present invention, it is preferable that the above-mentioned line width is obtained as one line width at the portion of the other value existing between one of the above-mentioned two values and the other value.

In practicing the present invention, it is preferable that the above-mentioned number of lines is obtained as a single number of lines of the other value existing between one of the above-mentioned two values and the other of the two values.

Further, in carrying out the present invention, the above-mentioned number of lines is one when the number of the other value existing between the one value of the above-mentioned two values and the one value is within the predetermined range. It may be obtained as the number of lines.

(Operation) According to such a configuration, the superposition positions of the collation fingerprint and the registered fingerprint are the line number distribution and the maximum line width distribution extracted from the respective binary graphic patterns stored in the collation and registered pattern memory. It is decided based on. By doing this, for example, even if some of the lines that make up the striped (line pattern) figure are dented inside the figure, the number of lines and the line width can be confirmed. Also, even if there is a break in a part of the line, that part has a small effect on determining the overlapping position, so that both patterns are overlapped quickly and with high accuracy, After that, these collations are performed.

(Embodiment) An embodiment of the fingerprint collation device of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing an embodiment of a fingerprint collation device of the present invention.

In FIG. 1, reference numeral 11 denotes a reading unit. The reading unit 11 optically reads a fingerprint to be collated and reads it as a binary graphic pattern (this is referred to as a collation pattern). For example, the figure is converted into a binary electric signal (also referred to as a black and white pattern for convenience of description) and output to the matching pattern memory 12.

Further, this collation pattern memory 12 stores this black-and-white pattern, and although not shown in the drawing, in this case, the addresses in the first direction (X direction) and the second direction (Y direction) of the memory are respectively designated. Of the first and second address counters.

Reference numeral 13 denotes a registered pattern memory for storing a fingerprint registered in advance, which has the same structure as the matching pattern memory 12.

Reference numeral 14 denotes a line number distribution extracting unit, and in the case of this embodiment, it is constituted by a first line number distribution extracting means 14a and a second line number distribution extracting means 14b, and also a first and second line number distribution. Extraction means 14a, 1
Although not shown, 4b includes a black dot line width counter, an X-direction line number distribution counter XHist, and a Y-direction line number distribution counter YHist (details will be described later). In the case of this embodiment, the first line number distribution extracting means 14a scans the collation pattern memory 12 in the raster scanning system in each of the X and Y directions, that is, the collation pattern memory 12
The matching patterns stored in the X- and Y-directions are read line-sequentially, line segments are counted by a method described later to obtain the number of lines for each scan, and thus the number-of-lines distribution in the X and Y directions is obtained. To create. On the other hand, the second line number distribution extracting means 14b scans the registered pattern memory 12 in each of the X direction and the Y direction by a raster scanning method, and similarly to the case where the line number distribution of the matching pattern is created by the first extracting unit 14a. A line number distribution for each X and Y direction of the registered pattern is created.

Reference numeral 15 denotes a line width distribution extraction unit, and in the case of this embodiment, it is constituted by a first line width distribution extraction means 15a and a second line width distribution extraction means 15b, and also a first and second line width distribution Extraction means 15a, 1
Although not shown in the figure, 5b is a black dot line width counter, an X-direction line width distribution register XWDT, and a Y-direction line width distribution register YW.
DT and line width maximum value register BMAX are provided respectively (details will be described later). And in the case of this embodiment,
The first line width distribution extraction means 15a stores X in the matching pattern memory 12.
And the Y-direction is scanned by the raster scanning method, that is, the matching pattern stored in the matching pattern memory 12 is line-sequentially read in the X-direction and the Y-direction, and the line segments on the scanning line are scanned by the method described later. The width is counted to find the maximum line width for each scan, thus creating a line width distribution for each of the X and Y directions. On the other hand, the second line width distribution extracting means 15b scans the registered pattern memory 13 in each of the X and Y directions by the raster scanning method, and the first line width distribution extracting means 15a creates the line width distribution of the matching pattern. Then, a line width distribution for each of the X and Y directions of the registered pattern is created.

Reference numeral 16 denotes a reference point extraction unit, which is a line number distribution extraction unit 14, in this case the first and second line number distribution extraction unit 14
comparing line number distributions of the matching pattern and the registered pattern respectively created in a and 14b for each of the X and Y directions, and a line width distribution extraction unit 15, in this case, first and second line width distribution extraction The line width distributions of the matching pattern and the registered pattern created by the sections 15a and 15b are X and Y, respectively.
A comparison is made for each direction to obtain the degree of difference between the two patterns, and a reference point for matching and matching the registered patterns is extracted based on the degree of difference.

Reference numeral 17 denotes a coincidence determination unit, which superimposes the matching pattern and the registered pattern respectively stored in the matching pattern memory 12 and the registered pattern memory 13 based on the reference points obtained by the reference point extracting unit 16, and matches both patterns. The degree is extracted by, for example, a method to be described later, and it is determined whether or not both patterns are fingerprints obtained from the same finger based on the value of the degree of coincidence.

Description of Line Number Distribution Extraction Unit Next, the operation of the line number distribution extraction unit will be described. In the case of this embodiment, the line number distribution extracting section is described as an example in which the first and second line number distribution extracting means 14a and 14b are configured. Has the same configuration except that the other is a registered pattern memory, the operation of the first line number distribution extracting means 14a will be described with reference to the flowchart shown in FIG. The description of the distribution extracting means 14b is omitted.

When scanning the matching pattern memory 12, first, the third
As shown in FIG. 3A, an example will be described in which the number of lines for each X coordinate of the matching pattern memory 12 is counted with the Y direction as the main scanning direction and the X direction as the sub scanning direction.

An initial value 0 is set in the X-direction line number distribution counter XHist provided in the first line number distribution extracting means 14a (step 21).

Sets the X-direction starting point coordinate values X _S of the collation pattern memory 12 to the first address counter X collation pattern memory 12 (step 22).

The black dot line width counter (sometimes abbreviated as CNT) provided in the first line number distribution extracting means 14a is set to 0 (step 23).

Setting the Y-direction starting point coordinate values Y _S matching pattern memory 12 to the second address counter Y collation pattern memory 12 (step 24).

After making such initial settings, the verification pattern memory 12
Reading the binary signal stored at the point (X _S , Y _S ),
If the signal is a black dot, 1 is added to CNT (steps 25 and 26). If the point (X _S , Y _S ) is not a black point and the value of CNT up to the present is not 0, XHist (X)
In this case, 1 is added to XHist (X _S ) (steps 25, 2
7) Finally, CNT is set to 0 (steps 25, 27, 2).
8, 29). Next, add 1 to the coordinate Y _S (step 30)
Then, the processes of steps 25 to 31 are repeated until the value of the address counter in the Y direction exceeds the end point coordinate Y _E (steps 30 and 31). That is, in this example, when scanning a matching pattern memory 12 in the Y-direction focusing on the coordinate X _S, CNT is incremented by 1 if the black spot appeared after the white point, if subsequently black spot is continuous Repeat the addition of 1 to count the number of black dots. The continuous portion of the black dots is determined to be a part of the line forming the collation pattern, and if a white dot appears again on the main scan, it is determined that the line segment ends and CNT is initialized to 0. And XHist (X _S ), which means that one line was counted
Is incremented by 1. The handle can be extracted the number of lines in the scan line in the Y direction of the coordinate X _S by performing a scanning line of the coordinate X _S.

Next, when the value of Y coordinate exceeds Y _E , CNT up to the present
If the value is not 0, 1 is added to XHist (X _S ) and 1 is added to the coordinate value X _S , and the processing of steps 23 to 35 is repeated until the end point coordinate X _E in the X direction is exceeded (step 34). ，
35).

By performing such processing, the number of lines after scanning in the Y direction for each X coordinate can be obtained, and therefore, the distribution of the number of lines in the X direction can be extracted.

Subsequently, as shown in FIG. 3 (B), the X-direction is the main scanning direction and the Y-direction is the sub-scanning direction.
The number of lines for each of the 12 Y coordinates is counted to extract the line number distribution in the Y direction (steps 41 to 55). This processing can be performed basically in the same procedure as that for obtaining the X-direction line number distribution described above, and therefore the description thereof will be omitted.

Using the second line number distribution extracting means 14b, the line number distribution of the registered pattern can be obtained in the same manner as described above.

Description of Line Width Distribution Extraction Unit Next, the operation of the line width distribution extraction unit will be described. In the case of this embodiment, the line width distribution extraction unit is described as an example in which the first and second line width distribution extraction means 15a and 15b are configured. Has the same configuration except that the other is a registered pattern memory, the operation of the first line width distribution extraction means 15a will be described with reference to the flowchart shown in FIG. The description of the distribution extraction means 15b is omitted.

When scanning the matching pattern memory 12, first, the third
As shown in FIG. 7A, an example will be described in which the maximum line width for each X coordinate of the matching pattern memory 12 is counted with the Y direction as the main scanning direction and the X direction as the sub scanning direction.

Sets the X-direction starting point coordinate values X _S of the collation pattern memory 12 to the first address counter X collation pattern memory 12 (step 61).

The black dot line width counter (C
Sometimes abbreviated as NT. ) Is set to 0 (step 62).

Initial value 0 in the line width maximum value register BMAX that stores the maximum value of the line width that meets on the scanning line when one scan is performed in the Y direction.
Is set (step 63).

Setting the Y-direction starting point coordinate values Y _S of the collation pattern memory 12 (step 64).

After making such initial settings, the verification pattern memory 12
Reading the binary signal stored at the point (X _S , Y _S ),
If the signal is not a black point, CNT is set to 0 (steps 65 and 66), and if the signal is a black point, 1 is added to CNT, and then the value of CNT is BMAX.
If it is larger than the value stored in, the value of CNT is stored in BMAX (steps 65, 67 to 69). Then, 1 is added to the coordinate Y _S (step 70), repeats the processing of step 65 to 71 until the value of the address counter in order Y-direction exceeds the end point coordinate Y _E (step 70,7
1). That is, in this example, when scanning a collation pattern memory 12 in the Y-direction focusing on the coordinate X _S, CNT is incremented by 1 if the black spot appeared after the white point, if subsequently black spot is continuous Repeat the addition of 1 to count the number of black dots. The continuous portion of the black dots is determined to be the line width of the line forming the matching pattern, and if a white dot appears again on the main scan, it is determined to be the end of the line segment. Then, the one having the largest number of black dots is extracted as the maximum line width in this scan from the group of black dots appearing in the Y-direction scan of the coordinate X _S.

Next, when the Y coordinate exceeds Y _E , the address XWDT (X) corresponding to the X coordinate of the X-direction line width distribution register XWDT, in this case, XWDT (X _S ) is determined by one scan as described above. The content of BMAX, which is the maximum value of the line width of the line, is stored (steps 71 and 72).

Then, 1 is added to the coordinate values X _S (step 73) to repeatedly perform the processing of steps 62-74 to over X-direction of the end point coordinates X _E of the collation pattern memory (step 73, 74).

By performing such processing, the maximum line width of the line segments that appear in the Y direction for each X coordinate can be obtained, and thus the line width distribution in the X direction can be extracted.

Subsequently, as shown in FIG. 3 (B), the X-direction is the main scanning direction and the Y-direction is the sub-scanning direction.
The line width for each of the 12 Y coordinates is counted to extract the line width distribution in the Y direction (steps 81 to 94). This processing can be performed by basically the same procedure as the above-described calculation of the X-direction line width distribution, and therefore the description thereof will be omitted.

Further, the line width distribution of the registered pattern can be obtained by using the second line width distribution extracting means 15b in the same manner as described above.

In description reference point extraction unit 16 of the reference point extraction unit, first, the X direction line number distribution XHist _t registration pattern, the X-direction line number distribution of the collation pattern XHist _t
Comparing the bets, and the X-direction line width distribution XWDT _t registration pattern, performed by comparing the X-direction line width distribution XWDT _s matching pattern overlaps best when superposed both patterns Find the position.

That is, when the matching pattern is moved by ΔX in the X direction, the line number distribution is XHist _t , the line width distribution is XWDT _t, and the difference between the line number and the line width distribution between the registered pattern and the matching pattern is β _ΔX . At this time, the difference β _ΔX can be _calculated, for example, by the following formula. From determined, the matching pattern [Delta] X values of beta _{[Delta] X} in the range of -LT ≦ ΔX ≦ LT is minimized, and the movement amount D _X in the X direction necessary for superposition with the registered pattern. In addition, this L
T may be set to a value larger than the maximum value of the deviation amount between the registered pattern and the matching pattern.

Next, the Y-direction line number distribution of the registered pattern is YHist _t , the line width distribution is YWDT _t , the Y-direction line number distribution of the matching pattern is YHist _S, and the line width distribution is YWDT _S. When the difference in the number of lines and the line width distribution is β _ΔY , this difference β _ΔY Then, ΔY that minimizes the value of β _ΔY within the range of −LT ≦ ΔY ≦ LT is set as the movement amount DY in the Y direction required for superposing the matching pattern with the registered pattern.

In this way, it is possible to extract a reference point at which the matching pattern and the registered pattern are superposed.

Description of Matching Judgment Unit The matching judgment unit 17 matches the matching pattern and the registered pattern when the matching pattern is moved by the overlapping movement amounts DX and DY obtained in the reference point extracting unit 16 in the X and Y directions, respectively. Is examined for all points (addresses) of the pattern memory to obtain the degree of coincidence α. The degree of coincidence α can be obtained by the following formula, for example.

Here, TP (X, Y) is the point (X,
Y) and SP (X, Y) represent the value of the point (X, Y) of the matching pattern memory 12, respectively. For example, when the point is a black point, it has a value of 1, and when it is a white point, it has a value of 0. It has a value of. The symbol Λ represents a logical product and the symbol V represents a logical sum.

The coincidence determination unit 17 further determines that the degree of coincidence α is a predetermined threshold T _a
A registration pattern when it is above, it is determined that they coincide with the collation pattern, also, both patterns when α is less than T _a is determined to be mismatched.

In the operation of the coincidence determination unit 17 described above, there is no inclination difference between the registered pattern and the matching pattern, and if the matching patterns are moved by the movement amounts DX and DY obtained by the reference point extracting unit 15. The example has been described in which the matching pattern completely overlaps the recording pattern. However, when actually performing matching, it is not possible to perform superimposition only by such simple movement, and it is necessary to perform fine adjustment in consideration of the inclination and the movement amount. When the degree of coincidence in such a case is γ, this γ can be obtained by the following formula, for example.

Note that RX = [(XX _o ) cos θ + (YY _o ) sin θ] RY = [(YY _o ) cos θ + (XX _o ) sin θ]. Here, x and y represent fine adjustment movement amounts in the X and Y directions, X _o and Y _o represent X and Y coordinates of the rotation center of the registered pattern, and θ represents a rotation angle.
Also, [] represents integerization.

Then,-? _K ??? _k , -LG? X? + LG, -L.
Seeking G ≦ y ≦ + LG maximum gamma _x in gamma in the range of, determines the value of the gamma _x matches the registration pattern and the collation pattern equal to or more than the above-mentioned T _a, also, gamma _x is T if it is less than _a two patterns is determined to match.

By the way, the value of θ _k may be an allowable value of the inclination angle difference between the registered pattern and the matching pattern, and when all angles are allowed, θ _k = 180 (degrees).

The value of LG may be the maximum extraction error of the movement amounts DX and DY obtained by the reference point extraction unit 15, for example.

The value of the registered pattern rotation center (X _o , Y _o ) may be the center of gravity of the registered pattern or the midpoint of a circumscribed quadrangle surrounding the registered pattern.

As described above, according to the fingerprint collating apparatus of the present invention, the fingerprint patterns to be collated can be quickly superposed at a position where both patterns substantially overlap each other, so that the fingerprint collating time can be shortened.

By the way, regardless of the magnitude of the minimum values of the differences β _ΔX and β _ΔY when comparing the line number distribution and line width distribution of the registered pattern with the line number distribution and line width distribution of the matching pattern in the above-described embodiment. Although an example of obtaining the degree of coincidence α by superimposing the two patterns has been described, it is also possible to immediately determine that the two patterns do not coincide when the minimum value exceeds the constant value T _b . By doing so, the time required for collation can be further shortened.

In the above-described embodiment, the line number distribution and the line width distribution are described as an example of extracting the patterns from the registered pattern and the matching pattern, respectively. However, regarding the line number distribution and the line width distribution of the registered pattern, the registered pattern is Even when the line number distribution and the line width distribution of this pattern are extracted at the time of registration and the line number and line width distribution are registered together with the registered pattern data, the same effect as the embodiment can be obtained. Can be done.

In the operation of the line number distribution extraction unit 14 of the above-described embodiment, X
The number of lines is judged to be present regardless of the size of the line width during scanning in the Y direction or the Y direction, that is, regardless of the number of black dots existing between white dots To 1
Is being added. However, if, for example, dust is read by the reading unit 11 and this dust is stored in the pattern memory as, for example, one black dot, on the contrary, the line of the figure is scanned in the longitudinal direction of the line, and a large number of black dots continue. Such cases are also possible. In consideration of such a situation, the line width may be counted as one line only when the line width is within a predetermined value range.
That is, in the case of the embodiment, in steps 28, 33, 48 and 53 of the flowchart shown in FIG. 2, the number of black dots existing between the white dots (the value of the black dot line width counter CNT).
Regardless of the size, the black spots disappear and white spots appear, and XHi
I added 1 to st or YHist, but the CNT value is W ₁
XHist or YHi only when ≦ CNT ≦ W ₂ is satisfied
By adding 1 to st, it is possible to deal with the situation that occurs when dust or a line is scanned.

Note that W ₁ and W ₂ are predetermined numbers, and the width of the convex portion (ridge) of the fingerprint is generally 0.2 to 0.5 mm.
If the reading resolution of the reading unit 11 is 20 pixels / mm, then if the width of the ridge is 0.5 mm, then 10 pixels,
Since 0.2 mm corresponds to 4 pixels, W ₁ = 4 and W ₂ = 10 are suitable for ridge determination.

In the operation of the line number distribution extracting unit 14 and the line width distribution extracting unit 15 of the above-described embodiment, the pattern memories 12 and 13 are set to X,
An example of obtaining the number of lines and the line width distribution by scanning along the Cartesian coordinate system of Y has been described, but the pattern memory has an X _D axis inclined by θ ₁ degrees with respect to the X axis and a θ ₂ degree with the Y axis. Even when scanning is performed along the tilted Y _D axis and the other processes are performed in the same manner as in the example to obtain the line number and line width distribution for each X _D and Y _D ,
The same effect as the embodiment can be obtained. In this case, it is preferable that the X _D and Y _D axes are not parallel.

Furthermore, in the operation of the line number distribution extracting unit 14 and the line width distribution extracting unit 15 of the above-described embodiment, an example of extracting the line number and the line width distribution by paying attention to the black points existing between the white points However, the present invention is not limited to this embodiment, for example, paying attention to the white point existing between the black point and the black point, this white point part as one line segment, other The number of lines and the line width distribution can be obtained by performing the same processing as in the embodiment.

Further, in the above-described embodiment, the line number distribution extracting section is configured by the first and second line number distribution extracting means, and the line width distribution extracting section is configured by the first and second line width distribution extracting means. Although the respective configurations have been described, the configurations of the number of lines and the line width distribution extraction unit are not limited to this embodiment, and can be changed according to the design. For example, time division may be performed, and the line number and line width distribution of the matching and registered patterns may be created by one line number and line width extracting means, respectively.

Further, it is obvious that the fingerprint collation device of the present invention can be applied to a collation device for figures other than fingerprints.

(Effects of the Invention) As is apparent from the above description, according to the fingerprint collation apparatus of the present invention, the line number distribution and the line width distribution are respectively obtained from the two fingerprint patterns to be collated in each of the first and second directions. It is provided with means for extracting and comparing these line number distributions and line width distributions in the first and second directions, respectively, and detecting the optimum position of superimposition of the two fingerprint patterns. Therefore, the superposition position of the two fingerprints can be accurately and quickly determined, and the time required for the collation can be shortened.

Therefore, it is possible to provide a fingerprint collation device that can collate fingerprints with high alignment accuracy.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of a fingerprint collation device of the present invention, FIG. 2 is a flow chart showing the operation of a line number distribution extracting section according to the present invention, and FIGS. 3 (A) and 3 (B) are the present invention. And FIG. 4 is a flow chart showing the operation of the line width distribution extraction unit according to the present invention, for explaining the line number distribution extraction unit and line width distribution extraction unit according to FIG. 11 …… reading section, 12 …… matching pattern memory 13 …… registered pattern memory 14 …… line number distribution extracting section 14a …… first line number distribution extracting means 14b …… second line number distribution extracting means 15 …… line Width distribution extraction unit 15a ... first line width distribution extraction means 15b ... second line width distribution extraction means 16 ... reference point extraction unit 17 ... match determination unit.

Claims (5)

[Claims] 1. A reading unit for reading a fingerprint to be collated as a binary collation pattern, a collation pattern memory for storing the collation pattern, and a registered pattern memory for storing a fingerprint registered in advance as the binary registration pattern. And scanning the collation pattern memory and the registered pattern memory in a raster system in each of the first direction and the second direction to obtain the line numbers of both patterns for each scan, and A line number distribution extraction unit that extracts the line number distribution for each second direction, and scans the matching pattern memory and the registered pattern memory in a raster method in each of the first direction and the second direction, and the maximum line of both patterns. Obtaining the width for each scan, a line width distribution extraction unit for extracting the line width distribution for each of the first and second directions of both patterns, and for each of the first and second directions A reference point extracting unit that compares the line width distributions and line number distributions of the matching pattern and the registered pattern with each other to extract the overlapping position of both patterns, and the matching and registration pattern based on the overlapping position. A fingerprint collation device, comprising: a coincidence determination unit that determines the degree of coincidence between both superposed patterns and determines whether or not the patterns coincide with each other based on the degree of coincidence. 2. The minimum difference obtained by comparing the line number distributions in the first direction and the line width distributions in the first direction of the two patterns with each other, and the second direction of the two patterns. Claims characterized in that the two patterns are judged as non-coincidence when at least one of the minimum dissimilarity obtained by comparing the line number distributions and the line width distributions is larger than a predetermined value. The fingerprint collation device according to item 1. 3. The line width according to claim 1, wherein a part of the other value existing between one of the two values is calculated as one line width. Alternatively, the fingerprint collation device according to item 2. 4. The line number is obtained by determining the portion of the other value existing between one value of the two values as one line number. The fingerprint collation device according to item 2 or 3. 5. The number of lines is obtained as one number of lines when the number of the other values existing between one of the two values is within a predetermined range. The fingerprint collation device according to claim 1 or 2, characterized in that.