JP7239034B2 - Fingerprint processing device, fingerprint processing method, program, fingerprint processing circuit - Google Patents

Description

The present invention relates to a fingerprint processing device, a fingerprint processing method, a program, and a fingerprint processing circuit.

When performing fingerprint matching, for example, the center of the fingerprint (core) is calculated from the fingerprint image, and information such as the position of the minutiae with respect to the center is used to compare the image of the fingerprint matching source and the fingerprint matching target. doing collation. However, if the position of the center of the fingerprint in the fingerprint image is specified incorrectly, the relationship between the center of the fingerprint and each minutiae is deviated, thereby degrading the matching accuracy. Patent Documents 1 to 3 disclose techniques related to fingerprint collation.

JP-A-10-177650 JP 2015-228070 A Japanese Patent No. 5822303

Due to the above-described problems, there is a demand for a fingerprint processing apparatus that performs fingerprint collation without specifying the center of the fingerprint and has a high collation speed.

SUMMARY OF THE INVENTION Accordingly, an object of the present invention is to provide a fingerprint processing device, a fingerprint processing method, a program, and a fingerprint processing circuit that solve the above problems.

According to the first aspect of the present invention, the fingerprint processing device, among the plurality of minutiae of the fingerprint specified in the fingerprint image to be collated, has the smallest first distance between each of the minutiae and other minutiae. A collation processing unit that determines a predetermined number of minutiae points selected in descending order of representative values based on the first degree of distance, as representative minutiae points to be used in the fingerprint collation processing.

According to the second aspect of the present invention, the fingerprint processing method includes, among a plurality of minutiae points of a fingerprint specified in a fingerprint image to be collated, a minimum first distance between each minutiae point and other minutiae points. A predetermined number of minutiae points selected in descending order of the representative value based on the first match indicating the distance are determined as the representative minutiae points to be used in the fingerprint matching process.

According to the third aspect of the present invention, the program instructs the computer, among the plurality of minutiae of the fingerprint specified in the fingerprint image to be collated, the smallest of the first distances between the other minutiae and the other minutiae. A predetermined number of minutiae points selected in descending order of the representative value based on the first match indicating the distance of the fingerprint are determined as the representative minutiae points to be used in the fingerprint matching process.

According to the fourth aspect of the present invention, the fingerprint processing circuit, among the plurality of minutiae of the fingerprint specified in the fingerprint image to be collated, minimizes each of the first distances between the minutiae and other minutiae. A predetermined number of minutiae points selected in descending order of the representative value based on the first match indicating the distance are determined as the representative minutiae points to be used in the fingerprint matching process.

According to the present invention, it is possible to provide a fingerprint processing device, a fingerprint processing method, a program, and a fingerprint processing circuit that enable high-speed collation without specifying the center of a fingerprint.

1 shows a fingerprint matching system including a fingerprint processing device; FIG. 2 is a hardware configuration diagram of a fingerprint processing device; FIG. 3 is a functional block diagram of the fingerprint processing device; FIG. It is a figure which shows the processing flow of a fingerprint processing apparatus. It is a figure for demonstrating the feature point which a feature point detection part detects. FIG. 3 is a diagram showing determined clear and blurred regions and minutiae in a fingerprint image; FIG. 4 is a diagram showing the stability of ridges determined in a fingerprint image; It is a figure which shows a representative degree calculation flow. FIG. 3 is a diagram visually showing feature points and representative feature points; FIG. 11 is a first diagram showing an overview of specifying feature amounts of representative feature points; FIG. 10 is a first diagram showing a processing flow for assigning feature amounts of representative feature points; FIG. 10 is a first diagram showing an outline of similarity calculation based on fingerprint feature quantities; FIG. 11 is a second diagram showing an overview of specifying feature amounts of representative feature points; FIG. 11 is a third diagram showing an overview of identifying feature amounts of representative feature points; FIG. 10 is a second diagram showing a processing flow for assigning feature amounts of representative feature points; FIG. 11 is a second diagram showing an outline of similarity calculation based on the fingerprint feature amount; 1 is a diagram showing the minimum configuration of a fingerprint processing device; FIG.

A fingerprint processing apparatus according to an embodiment of the present invention will now be described with reference to the drawings.
FIG. 1 is a diagram showing a fingerprint matching system including a fingerprint processing device according to the same embodiment.
The fingerprint collation system includes a fingerprint processor 1, a fingerprint reader 2, and a database 3, as shown in FIG. A fingerprint processing device 1 is connected to a fingerprint reader 2 via a communication cable. The fingerprint processing device 1 is also connected to the database 3 via a communication cable. The fingerprint processing device 1 compares the fingerprint information obtained from the fingerprint image obtained from the fingerprint reader 2 with the fingerprint information stored in the database 3 to perform fingerprint collation processing. Fingerprint information obtained from fingerprint images of many people is recorded in the database 3 in advance.

FIG. 2 is a hardware configuration diagram of the fingerprint processing device.
The fingerprint processing device 1 includes a CPU (Central Processing Unit) 11, a RAM (Random Access Memory) 12, a ROM (Read Only Memory) 13, an SSD (Solid State Drive) 14, a communication module 15, a display screen 16, an IF (Interface). 17 and the like. The fingerprint processing device 1 is a computer having such functions.

FIG. 3 is a functional block diagram of the fingerprint processing device.
Fingerprint processing device 1 executes a fingerprint processing program in CPU 11 . Thus, the fingerprint processing apparatus 1 has the functions of an image acquisition section 21 , a sharpness determination section 22 , a ridge quality determination section 23 , a feature point detection section 24 and a collation processing section 25 . The fingerprint processing device 1 also includes a storage unit 15 .

An image acquisition unit 21 acquires a fingerprint image from the fingerprint reader 2 .
The sharpness determination unit 22 determines and classifies the fingerprint sharpness of each pixel in the fingerprint image.
A ridge quality determination unit 23 determines the quality of the ridges of the fingerprint appearing in each pixel in the fingerprint image.
The minutiae detection unit 24 detects the minutiae of the fingerprint appearing in the fingerprint image.
The collation processing unit 25 performs collation processing by comparing the fingerprint information obtained from the fingerprint image acquired by the image acquisition unit 21 and the fingerprint information of a plurality of persons recorded in the database 3 .

In the present embodiment, the matching processing unit 25 performs the first match based on the first match indicating that the distance between the fingerprint minutiae identified in the fingerprint image to be matched and the other minutiae is long. A plurality of minutiae having a large joint value are determined to be representative minutiae used for fingerprint collation.
Alternatively, the collation processing unit 25 determines, among the fingerprint minutiae identified in the fingerprint image to be collated, the second degree indicating that the distance to the pixel area where the sharpness of the image identified in the fingerprint image is unknown is long. A plurality of feature points with large second degree values are determined as representative feature points.
Alternatively, the collation processing unit 25 indicates that the first match indicates that the distance between the other feature points is long, and that the distance to the pixel area where the sharpness of the image specified in the fingerprint image is unknown is long. A plurality of representative feature points are determined using the second degree. For example, the first degree may be multiplied by the second degree, the representativeness may be calculated based on this value, and a plurality of representative feature points may be determined based on the representativeness.

The collation processing unit 25 may further use the degree of stability of ridges in the pixels corresponding to the specified fingerprint minutiae in the fingerprint image to calculate the degree of representativeness, and determine the representative minutiae based on the degree of representativeness.

Then, the collation processing unit 25 identifies a plurality of representative minutiae representing features with a high degree of representation, and identifies a plurality of representative minutiae in the fingerprint image to be collated and the fingerprint image to be collated recorded in the database 3. A matching process is performed using a plurality of representative feature points.

FIG. 4 is a diagram showing the processing flow of the fingerprint processing device.
Next, the details of the processing of the fingerprint processing device will be explained step by step.
First, the image acquisition unit 21 of the fingerprint processing device 1 acquires a fingerprint image from the fingerprint reader 2 (step S101). The image acquisition unit 21 temporarily records the fingerprint image in the RAM 12 or the like. Then, the sharpness determination unit 22 reads the fingerprint image from the RAM 12 and determines the sharpness of each rectangular area set over the entire fingerprint image (step S102). Each rectangular area may be an area corresponding to a pixel, or may be a rectangular area set separately from the pixel. After determining the definition of each rectangular area, the definition determining unit 22 records an ID specifying the rectangular area in the fingerprint image and the definition in the RAM 12 . Sharpness may be binary, for example sharp or blurry. Any method may be used to determine the sharpness. For example, the sharpness determination section 22 may calculate the sharpness by the same process as the process performed by the quality index extraction means described in Japanese Patent Application Laid-Open No. 10-177650.

The ridge quality determination unit 23 reads the fingerprint image from the RAM 12, specifies pixels in which ridges appear in the above-described rectangular area in the fingerprint image, and determines the stability of the ridges at those pixels (step S103). The ridge stability means the stability of the ridge width. The ridge stability determined by the ridge quality determining unit 23 is indicated by variance values of both a black image showing fingerprint ridges and a white image showing spaces between fingerprint ridges in a set rectangular area, for example. The ridge quality judging unit 23 judges, as a rectangular area with high ridge stability, a rectangular area in which both the black image representing the fingerprint ridges and the white image representing the space between the fingerprint ridges have variance values higher than a threshold value. On the other hand, the ridge quality judging unit 23 judges a rectangular area in which the variance value of either the black image representing the fingerprint ridges or the white image representing the space between the fingerprint ridges is equal to or less than the threshold as a rectangular area with low ridge stability. do. The ridge quality determination unit 23 records information in which the ID of the rectangular area is associated with a flag indicating whether the stability of the ridge is high or low in the RAM 12 or the like. More detailed calculation of the ridge line stability may be performed using, for example, the technique described in steps S301 to S305 of Japanese Patent No. 5822303, or the like.

The minutiae detection unit 24 reads the fingerprint image from the RAM 12 and detects minutiae of the fingerprint appearing in the fingerprint image (step S104). A conventional detection method may be used to detect feature points. More specifically, the feature point detection unit 24 extracts two types of feature points from the fingerprint image.

FIG. 5 is a diagram for explaining feature points detected by the feature point detection unit.
The first feature point targeted by the feature point detection unit 24 is the end point (end point) of the ridge. FIG. 5A shows a first feature point P1 sandwiched between two ridges 501 and 502. FIG. The minutiae detection unit 24 detects the end points of ridges such as the first minutiae P1 shown in FIG. 5A as the first minutiae P1 of the input fingerprint image. Specifically, the feature point detection unit 24 detects the first feature points P1 while scanning the entire fingerprint image.

The second feature points targeted by the feature point detection unit 24 are ridge bifurcation points. FIG. 5(b) shows a second feature point P2 where the ridge 511 branches into two ridges 512 and 513. FIG. The minutiae detection unit 24 detects ridge branching points such as the second minutiae P2 shown in FIG. 5B as the second minutiae P2 of the input fingerprint image.

Next, the minutiae detection unit 24 calculates the extending direction of the ridge line of the fingerprint extending from each detected minutiae. As for the first feature point P1, as shown in FIG. 5A, the direction in which the ridge 500 is generated from the first feature point P1 (the direction of the arrow 504) is the extension of the ridge from the first feature point P1. Defined as direction.

Regarding the second feature point P2, as shown in FIG. The direction (direction of arrow 514) traveling between two ridges (the middle of the vector) is defined as the extension direction of the ridge from the second feature point P2.

The minutia detection unit 24 records in the RAM 12 the position information of the minutiae detected in the fingerprint image and the information on the extending direction of the ridge extending from the minutiae (stretching direction information). The position information (coordinates) of the detected first feature point P1 and second feature point P2 may be recorded in the form of pixel positions and rectangular area IDs in the fingerprint image. Also, the angle between the X-axis in the XY coordinate system with each feature point set as the origin and the extending direction of the ridge from the feature point may be recorded as the extending direction of the ridge from each feature point.

FIG. 6 is a diagram showing the determined clear and blurred regions and minutiae in a fingerprint image.
The shaded area in the fingerprint image shown in this figure indicates the blurred area E1 whose definition is less than the threshold. A non-shaded area indicates a clear area E2 whose sharpness is equal to or higher than the threshold. In such a fingerprint image, local points surrounded by circles and squares in the image indicate feature points. Among the feature points, the feature points indicated by circles are the first feature points P1. Among the feature points, the feature points indicated by square marks are the second feature points P2. The straight lines extending from the circles and squares indicate the extending directions of the ridges extending from the respective feature points.

FIG. 7 is a diagram showing the stability of ridges determined in a fingerprint image.
The fingerprint image shown in FIG. 7 shows a white area with a high ridge stability (quality) and a black area with a ridge stability lower than the threshold in the clear area E2. As a result of the processing in step S103, the fingerprint processing apparatus 1 may output an image showing the distribution of the stability of the ridges shown in FIG.

The collation processing unit 25 calculates the representativeness eval_p of each feature point using the position information and the extension direction information for each feature point detected by the feature point detection unit 24 (step S105). This representativeness is a frequency of the feature points identified in the fingerprint image that increases in value as the feature points are further away from other feature points. The representativeness eval_p may be a frequency indicating that the feature points identified in the fingerprint image are distant from other feature points and have high definition.

FIG. 8 is a diagram showing a representative degree calculation flow.
In calculating the representativeness eval_p, the collation processing unit 25 first acquires from the RAM 12 position information and extension direction information for each feature point detected by the feature point detection unit 24 . The matching processing unit 25 selects one selected feature point from among the feature points, and calculates the distances between the selected feature point and all other feature points (step S1051). The collation processing unit 25 determines the information of the smallest distance among the calculated distances as the distance Lmp (first match) to the closest feature point of the selected feature point (step S1052).

The collation processing unit 25 then uses the position information of the selected feature points and the position information of all the rectangular areas indicating the blurred area E1 to calculate their distances (step S1053). The collation processing unit 25 determines the information of the smallest distance among the calculated distances as the distance ucz (first match) to the rectangular area indicating the blurred area E1 closest to the selected feature point (step S1054). ). The collation processing unit 25 also acquires from the RAM 12 information on the ridge line stability mRql of the pixel corresponding to the selected feature point. Then, the representativeness eval_p of the selected feature point is calculated by Equation (1) (step S1055).

In formula (1), A and B represent constants. Moreover, "/" in Formula (1) means a division (÷). The constant A is used to adjust the value obtained by multiplying the distance Lmp to the closest feature point to the selected feature point by the distance ucz to the rectangular area indicating the blurred area E1 closest to the selected feature point so that the product does not become too large. is a constant of The constant B is a constant for adjusting the value of the ridge line stability mRql of the pixel corresponding to the selected feature point. Expression (1) can be used to calculate the representativeness eval_p that indicates whether the selected feature point has a high distance from other feature points and a high degree of definition. By calculating the representativeness eval_p using the distance ucz to the rectangular area indicating the blurred area E1 closest to the selected feature point, as in Equation (1), other feature points exist in the blurred area E1. It is possible to calculate the representativeness eval_p assuming the case. Although information about the stability mRql of the ridge of the pixel corresponding to the selected feature point is added in equation (1), the representativeness eval_p may be calculated without adding this information. In addition, in the equation (1), the information of the distance ucz to the rectangular area indicating the blurred area E1 closest to the selected feature point is added, but the representativeness eval_p may be calculated without adding this information. .

The collation processing unit 25 determines whether all the feature points appearing in the fingerprint image have been processed to calculate the representativeness eval_p as selected feature points (step S106). The collation processing unit 25 repeats the calculation of the representativeness eval_p for all the feature points appearing in the fingerprint image as the selected feature points until the calculation of the representativeness eval_p is completed. The collation processing unit 25 searches the calculated representativeness eval_p for all the feature points in descending order, and identifies a plurality of predetermined number of feature points having a large representativeness eval_p as representative feature points (step S107). The predetermined number may be a number such as 20 or 30, for example.

FIG. 9 is a diagram visually showing feature points and representative feature points.
As shown in FIG. 9, many feature points are detected in the fingerprint image. The collation processing unit 25 calculates the representativeness eval_p of each feature point, and determines a predetermined number of feature points selected in descending order as representative feature points. In FIG. 9, the feature points indicated by black marks have high representativeness eval_p, and the feature points indicated by light-colored marks have low representativeness eval_p.

FIG. 10 is a first diagram showing an overview of specifying feature amounts of representative feature points.
FIG. 11 is a first diagram showing a processing flow for assigning feature amounts of representative feature points.
The collation processing unit 25 then assigns feature amounts to each of the plurality of representative feature points (step S108). Specifically, the collation processing unit 25 selects one representative feature point from a plurality of representative feature points (step S108-1). This selected representative feature point is called a selected representative feature point. The collation processing unit 25 reads the position information and the extension direction information (representative extension direction) of the selected representative feature point from the RAM 12 . The collation processing unit 25 also reads from the RAM 12 position information and extension direction information of other feature points in the fingerprint image. The collation processing unit 25 assumes that the extension direction indicated by the extension direction information (representative extension direction) for the selected representative feature point is the positive direction of the vertical axis, and the vertical axis based on the extension direction information is centered on the selected representative feature point. , the divided areas of the first to eighth quadrants obtained by dividing the 360-degree direction on the fingerprint image plane at every 45-degree angle are specified (step S108-2).

It is now assumed that the collation processing unit 25 has set the representative feature point numbered 1 in FIG. 10 as the selected representative feature point. In this case, the axis extending in the extension direction indicated by the extension direction information with the first selected representative feature point as the origin in FIG. 10 is referred to as the axis a. An axis rotated 45 degrees to the left from the axis a about the first selected representative feature point is called an axis b. An axis rotated 45 degrees to the left from the axis b about the first selected representative feature point is called an axis c. An axis rotated 45 degrees to the left from the axis c about the first selected representative feature point is called an axis d. An axis rotated 45 degrees to the left from the axis d about the first selected representative feature point is called an axis e. An axis rotated 45 degrees to the left from the axis e about the first selected representative feature point is called an axis f. An axis rotated 45 degrees to the left from the axis f about the first selected representative feature point is called an axis g. An axis rotated 45 degrees to the left from the axis g about the first selected representative feature point is called an axis h.

The collation processing unit 25 identifies the 9th feature point closest to the 1st selected representative feature point in the first quadrant between the axis a and the axis b (step S108-3). The collation processing unit 25 also specifies the 17th feature point closest to the 1st selected representative feature point in the second quadrant between the axis b and the axis c. Similarly, the collation processing unit 25 specifies the feature point closest to the selected representative feature point No. 1 in the third to eighth quadrants divided by each axis. The collation processing unit 25 calculates the distance from the first selected representative feature point to each feature point specified in the first quadrant to the eighth quadrant, and the extension direction of each feature point specified in the first quadrant to the eighth quadrant. Based on all of the information and the representativeness calculated for the first selected representative feature point, the feature amount of the first selected representative feature point is determined (step S108-4). The collation processing unit 25 determines whether or not feature amounts have been identified for all representative feature points (step S108-5). If the matching processing unit 25 has not specified the feature amount for all the representative feature points, it repeats the process of specifying the feature amount.

After specifying the feature amounts for all the representative minutiae, the collation processing unit 25 determines the information consisting of the feature amounts for all the representative minutiae as the fingerprint feature amount for the fingerprint image acquired from the fingerprint reader 2. (step S109). A fingerprint image acquired from the fingerprint reader 2 is hereinafter referred to as an authentication fingerprint image. The collation processing unit 25 reads the fingerprint feature amount of the first fingerprint image recorded in the database 3 . A fingerprint image recorded in the database 3 is hereinafter referred to as a search target fingerprint image. The fingerprint feature amount of the search target fingerprint image is a feature amount calculated in the same manner as the fingerprint feature amount of the authentication fingerprint image. The collation processing unit 25 calculates the degree of similarity using the fingerprint feature amount of the authentication fingerprint image and the fingerprint feature amount of the search fingerprint image (step S110). Similarity calculation is one aspect of matching processing.

FIG. 12 is a first diagram showing an outline of similarity calculation based on the fingerprint feature amount.
As shown in FIG. 12, the collation processing unit 25 performs, for each of the plurality of representative minutiae specified in the authentication fingerprint image, the representativeness, the distance to the nearest The RAM 12 stores fingerprint feature amount information including extending direction information indicating the extending direction of the ridge extending from the nearest neighboring minutiae. A plurality of search target fingerprint images recorded in the database 3 are also composed of similar fingerprint feature amount information.

The collation processing unit 25 round-robinly calculates the difference between each representative minutiae on the authentication fingerprint image side and each representative minutiae point on the search target fingerprint image. When calculating the difference between one representative minutiae on the authentication fingerprint image side and one representative minutiae point on the search target fingerprint image, for the same corresponding quadrant, the difference in the distance to the nearest neighbor minutiae and the extension direction Each difference in the direction indicated by the information is multiplied by the representativeness of the authentication fingerprint image, and then added to calculate the value, and the cumulative value for eight quadrants of the value is calculated. This value is the difference between one representative minutiae on the authentication fingerprint image side and one representative minutiae on the search target fingerprint image.

Assuming that the representative minutiae specified in the authentication fingerprint image are 1 to n, and the representative minutiae specified in the search target fingerprint image are 1 to N, the matching processing unit 25 determines one representative minutiae in the authentication fingerprint image and the search destination. A difference from each representative feature point in the fingerprint image is calculated.
Assuming that one representative minutiae in the authentication fingerprint image is the representative minutiae 1, the collation processing unit 25 calculates the difference S( 11, 12, . . . 1N).
The collation processing unit 25 specifies the representative minutiae of the search target fingerprint image with the smallest difference S among the calculated differences S. FIG. Through this process, the collation processing unit 25 can identify representative minutiae included in the search fingerprint image that are close to one representative minutiae 1 identified in the authentication fingerprint image.
The collation processing unit 25 identifies a representative minutiae having a small difference S in the search target fingerprint image for all the representative minutiae (1 to n) in the authentication fingerprint image.
The matching processing unit 25 sorts the differences S between the representative minutiae (1 to n) in the authentication fingerprint image and the representative minutiae in the search target fingerprint image specified in ascending order, and sorts them by a certain number (2 to n) in descending order. 4) Calculate the average of the values. The collation processing unit 25 calculates the degree of similarity between the authentication fingerprint image and the search target fingerprint image by subtracting the average value from the constant. The value of the search fingerprint image with a high degree of similarity is high because the value of subtraction from the constant is small.

The collation processing unit 25 uses all search target fingerprint images contained in the database 3 to calculate the degree of similarity with the authentication fingerprint image. The collation processing unit 25 determines a search target fingerprint image whose similarity is higher than the reference threshold as a collation candidate for more detailed collation determination with the authentication fingerprint image (step S111). The matching processing unit 25 performs matching between the authentication fingerprint image and the search target fingerprint image determined as the matching candidate using the feature points included in each fingerprint image (step S112). For the matching process, a known technique of performing matching determination using feature points may be used. As a result of the matching determination, the matching processing unit 25 determines that the search target fingerprint image with the highest similarity among the search target fingerprint images with higher similarity than the reference value is the fingerprint image that matches the authentication fingerprint image, and outputs it. (Step S113). If the search target fingerprint image having a high degree of similarity cannot be retrieved as a result of the collation determination, the collation processing unit 25 outputs information indicating no match (step S114).

To briefly explain the above process, first, the collation processing unit 25 determines the representative extension direction of the fingerprint ridges extending from the extracted representative minutiae for one extracted representative minutiae extracted from a plurality of representative minutiae. Identify. In addition, the collation processing unit 25 determines the distance to the neighboring minutiae determined based on the extracted representative minutiae, the extending direction based on the representative extending direction of the ridges of the fingerprint extending from the neighboring minutiae, and the proximate minutiae. and the representativeness of the extracted representative feature points. Then, the collation processing unit 25 uses the feature values similarly assigned to each of the extracted representative minutiae to compare the plurality of representative minutiae in the fingerprint image of the collation source (authentication fingerprint image) with the fingerprint image of the collation destination (search destination). A fingerprint matching process is performed using a plurality of representative minutiae in a fingerprint image).

FIG. 13 is a second diagram showing an overview of specifying feature amounts of representative feature points.
FIG. 14 is a third diagram showing an overview of specifying feature amounts of representative feature points.
FIG. 15 is a second diagram showing a processing flow for assigning feature amounts of representative feature points.
The collation processing unit 25 may specify search target fingerprint images as collation candidates by methods other than the processes described with reference to FIGS. 10 to 12 .
The collation processing unit 25 assigns feature amounts to each of the plurality of representative feature points (step S108). Specifically, the collation processing unit 25 selects one representative feature point from a plurality of representative feature points (step S108-11). In FIG. 13, the representative feature point indicated by reference numeral 1 is the selected representative feature point. This selected representative feature point is called a selected representative feature point. The matching processing unit 25 reads from the RAM 12 the position information and the extending direction information about the selected representative feature point. The collation processing unit 25 reads from the RAM 12 the position information and extension direction information of other feature points (including feature points other than the representative feature points) in the fingerprint image. The collation processing unit 25 assumes that the extension direction indicated by the extension direction information for the selected representative feature point is the plus direction of the vertical axis, and determines the extension direction of the other feature points opposite to the extension direction of the vertical axis. A ridge extension direction determination image (FIG. 13) is generated in which neighboring pixels are white, and pixels in the vicinity of feature points whose extending directions are the same as the vertical axis are black (step S108). -12). For an image that is not identified as black or white in the ridge extension direction determination image, an interpolated color value is calculated based on the colors (pixel values) of other pixels in the vicinity, and a range from black to white is calculated. It may have a pixel value of any one of the gradation values. In FIG. 13, it can be seen that black pixels surround a feature point 50 whose extension direction is the same as the extension direction of the ridge of the selected representative feature point 1 . In addition, in FIG. 13, it can be seen that white pixels surround the feature points 9 and 6 whose extending directions are opposite to the extending direction of the ridge of the selected representative feature point 1 .

The collation processing unit 25 extracts pixel values of 24 feature extraction points centered on the position of the selected representative feature point 1 in the ridge extension direction determination image (FIG. 13) generated for the selected representative feature point 1 ( step S108-13). FIG. 14 shows the positions of 24 feature extraction points (1) to (24) with the position of the selected representative feature point 1 as the center. The collation processing unit 25 converts the value obtained by arranging the pixel values of the feature extraction points (1) to (24) and the information indicating the stability of the ridge calculated for the representative minutiae to the fingerprint feature of the selected representative minutiae 1. amount is determined (step S108-14). The collation processing unit 25 determines whether the fingerprint feature amount has been calculated for all the representative minutiae (step S108-15). If fingerprint feature quantities have not been calculated for all representative minutiae, the collation processing unit 25 generates fingerprint feature quantities for other representative minutiae by the same method.

After calculating the fingerprint feature amounts for all the representative minutiae specified in the authentication fingerprint image, the collation processing unit 25 converts the information consisting of the fingerprint feature amounts for all of the representative minutiae into the fingerprint feature amounts for the authentication fingerprint image. (step S109). The fingerprint feature amount of the search target fingerprint image is a feature amount calculated in the same manner as the fingerprint feature amount of the authentication fingerprint image. The relationship between the position of the representative minutiae and the positions of the feature extraction points (1) to (24) in the authentication fingerprint image, and the position of the representative minutiae and the positions of the feature extraction points (1) to (24) in the search target fingerprint image. is the same as the relationship with The collation processing unit 25 calculates the degree of similarity using the fingerprint feature amount of the authentication fingerprint image and the fingerprint feature amount of the search fingerprint image (step S110).

FIG. 16 is a second diagram showing an outline of similarity calculation based on the fingerprint feature amount.
As shown in this figure, the fingerprint feature quantity of the authentication fingerprint image includes the stability of ridges (ridge quality) in pixels (or rectangular areas) corresponding to a plurality of representative minutiae identified from many minutiae. , and the pixel values of the surrounding feature extraction points (1) to (24) based on the representative feature point. A plurality of search target fingerprint images recorded in the database 3 are also composed of similar fingerprint feature amount information.

The collation processing unit 25 round-robinly calculates the difference between each representative minutiae on the authentication fingerprint image side and each representative minutiae point on the search target fingerprint image. When calculating the difference S between one representative minutiae on the authentication fingerprint image side and one representative minutiae point on the search target fingerprint image, the difference between the pixel values of the corresponding feature extraction points is the ridge line of the authentication fingerprint image. is calculated for each of the 24 feature extraction points and accumulated. This value is the difference S between one representative minutiae on the authentication fingerprint image side and one representative minutiae on the search target fingerprint image.

Assuming that the representative minutiae specified in the authentication fingerprint image are 1 to n, and the representative minutiae specified in the search target fingerprint image are 1 to N, the matching processing unit 25 determines one representative minutiae in the authentication fingerprint image and the search destination. A difference S from each representative feature point in the fingerprint image is calculated.
Assuming that one representative minutiae in the authentication fingerprint image is the representative minutiae 1, the collation processing unit 25 calculates the difference S( 11, 12, . . . 1N).
The collation processing unit 25 specifies the representative minutiae of the search target fingerprint image with the smallest difference S among the calculated differences S. FIG. Through this process, the collation processing unit 25 can identify representative minutiae included in the search fingerprint image that are close to one representative minutiae 1 identified in the authentication fingerprint image.
The collation processing unit 25 identifies a representative minutiae having a small difference S in the search target fingerprint image for all the representative minutiae (1 to n) in the authentication fingerprint image.
The matching processing unit 25 sorts the differences S between the representative minutiae (1 to n) in the authentication fingerprint image and the representative minutiae in the search target fingerprint image specified in ascending order, and sorts them by a certain number (2 to n) in descending order. 4) Calculate the average of the values. The collation processing unit 25 calculates the degree of similarity between the authentication fingerprint image and the search target fingerprint image by subtracting the average value from the constant. The value of the search fingerprint image with a high degree of similarity is high because the value of subtraction from the constant is small.

The collation processing unit 25 uses all search target fingerprint images contained in the database 3 to calculate the degree of similarity with the authentication fingerprint image. The collation processing unit 25 determines a search target fingerprint image whose similarity is higher than the reference threshold as a collation candidate for more detailed collation determination with the authentication fingerprint image (step S111). The matching processing unit 25 performs matching between the authentication fingerprint image and the search target fingerprint image determined as the matching candidate using the feature points included in each fingerprint image. For the matching process, a known technique of performing matching determination using feature points may be used. As a result of the collation determination, the collation processing unit 25 determines that the search target fingerprint image with the highest degree of similarity is the fingerprint image that matches the authentication fingerprint image (step S112). If the search target fingerprint image having a high degree of similarity cannot be retrieved as a result of the collation determination, the collation processing unit 25 outputs information indicating no match (step S113).

To briefly explain the above process, first, the collation processing unit 25, for one extracted representative minutia extracted from a plurality of representative minutia, identify. Next, the collation processing unit 25 indicates that the extending direction based on the representative extending direction of the ridges of the fingerprint extending from the neighboring minutiae determined based on the extracted representative minutiae and the representative extending direction have a high degree of matching. A ridge extension direction determination image is generated that shows the pixel values of the neighboring feature points and the pixel values of the neighboring feature points indicating that the degree of matching with the representative extension direction is low. Then, the collation processing unit 25 assigns to the representative feature point a feature quantity indicating a plurality of pixel values at predetermined positions with reference to the position of the corresponding extracted representative feature point in the ridge extension direction determination image. The matching processing unit 25 uses the feature values similarly assigned to each of the representative minutiae to compare the plurality of representative minutiae in the fingerprint image to be matched and the plurality of representative minutiae in the fingerprint image to be matched. Perform matching process.

The embodiment of the present invention has been described above. According to the above-described processing, the matching process is performed without specifying the center of the fingerprint. be able to.
In addition, simple four arithmetic operations and sorting processing are used in the process of determining the search target fingerprint image of the matching candidate described above, and the feature points used for calculation are limited to a small number of feature points such as representative feature points and feature extraction points. . This makes it possible to reduce the processing for determining matching candidates.
Also, in the matching process, only the search target fingerprint image specified as a matching candidate is used to perform the matching process with the authentication fingerprint image, so the process can be reduced.

In the above description, the fingerprint processing apparatus 1 performs either the similarity calculation described with reference to FIGS. 10 to 12 or the similarity calculation described with reference to FIGS. 13 to 16. However, both similarity calculations may be performed to identify matching candidates with high similarity.

FIG. 17 is a diagram showing the minimum configuration of the fingerprint processing device.
As shown in this figure, the fingerprint processing apparatus 1 performs at least the first match indicating that the distance between the fingerprint minutiae specified in the fingerprint image to be collated and the other minutiae is long. It has a collation processing unit 25 for determining a plurality of minutiae having a large first match value as representative minutiae used for fingerprint collation.
The fingerprint processing apparatus 1 according to this embodiment may be defined as a verification processing circuit having at least the functions of the verification processing section 25 described above.

The fingerprint processing device 1 described above has a computer system inside. A program for causing the fingerprint processing device 1 to perform each of the processes described above is stored in a computer-readable recording medium of the device. The above processing is performed. Here, the computer-readable recording medium refers to magnetic disks, magneto-optical disks, CD-ROMs, DVD-ROMs, semiconductor memories, and the like. Alternatively, the computer program may be distributed to a computer via a communication line, and the computer receiving the distribution may execute the program.

Further, the program may be for implementing part of the functions of the processing units described above. Further, it may be a so-called difference file (difference program) that can realize the above-described functions in combination with a program already recorded in the computer system.

1 Fingerprint processing device 2 Fingerprint reader 3 Database 11 CPU
12 RAM
13 ROM
14 SSD
15... communication module 16... display screen 17... IF (interface)
21... Image acquisition unit 22... Sharpness determination unit 23... Ridge quality determination unit 24... Characteristic point detection unit 25... Verification processing unit

Claims (12)

Determining a representative minutiae to be used in the fingerprint matching process based on a first distance to a pixel region in the fingerprint image whose image sharpness is unknown among a plurality of minutiae of the fingerprint identified in the fingerprint image to be matched. Having a matching processing unit that
Fingerprint processor. The matching processing unit determines a representative minutiae to be used for the fingerprint matching process based on a second distance from other minutiae among the plurality of minutiae.
The fingerprint processing device according to claim 1. The matching processing unit determines, as the representative feature points, a predetermined number of feature points selected in descending order of the representative value based on the second degree indicating the minimum distance among the second distances. have
The fingerprint processing device according to claim 2. The collation processing unit calculates a representativeness based on a value obtained by multiplying the second degree by a first degree indicating that the first distance is long, and determines the representative feature point based on the representativeness. The fingerprint processing device according to claim 3. The collation processing unit further uses a degree of stability of ridges in pixels corresponding to the minutiae of the fingerprint identified in the fingerprint image to calculate a degree of representation, and determines the representative minutiae based on the degree of representation. 5. The fingerprint processing device according to any one of items 1 to 4. The collation processing unit specifies a plurality of representative minutiae representing features with a high degree of representation, and a plurality of representative minutiae points in the fingerprint image to be collated and a plurality of representative minutiae in the fingerprint image to be collated. 6. The fingerprint processing device according to claim 4, wherein the collation processing is performed using the . The collation processing unit specifies, for one extracted representative minutiae extracted from the plurality of representative minutiae, a representative extension direction of ridges of a fingerprint extending from the extracted representative minutiae, and based on the extracted representative minutiae, A feature amount determined based on the distance to the determined neighboring minutiae, the extending direction of the ridges of the fingerprint extending from the neighboring minutiae with reference to the representative extending direction, and the representativeness of the neighboring minutiae. , and using the feature quantity similarly assigned to each of the representative minutiae, the fingerprint of the plurality of representative minutiae in the fingerprint image of the matching source and the plurality of representative minutiae in the fingerprint image of the matching destination 7. The fingerprint processing device according to claim 6, which performs collation processing. The collation processing unit determines that the neighboring feature points are feature points in the vicinity of the extracted representative feature point in each of eight quadrant regions with the extracted representative feature point as the origin, with the representative stretching direction of the extracted representative feature point as a reference. The fingerprint processing device according to claim 7. The collation processing unit specifies, for one extracted representative minutiae extracted from the plurality of representative minutiae, a representative extension direction of ridges of a fingerprint extending from the extracted representative minutiae, and based on the extracted representative minutiae, a pixel value of the neighboring minutiae indicating that the extension direction of the ridges of the fingerprint extending from the determined neighboring minutiae based on the representative stretching direction and the representative stretching direction is high, and the representative stretching direction; A ridge extension direction determination image is generated that indicates the pixel values of the neighboring feature points that indicate a low degree of matching with the ridge extension direction determination image, and the position of the corresponding extracted representative feature point in the ridge extension direction determination image is used as a reference. A feature quantity indicating a plurality of pixel values at a predetermined position is assigned, and using the feature quantity similarly assigned to each of the representative feature points, the plurality of representative feature points in the fingerprint image of the reference source and the fingerprint of the reference target are obtained. 7. The fingerprint processing device according to claim 6, wherein fingerprint matching processing is performed using a plurality of representative feature points in an image. A fingerprint processing device
Determining a representative minutiae to be used in the fingerprint matching process based on a first distance to a pixel region in the fingerprint image whose image sharpness is unknown among a plurality of minutiae of the fingerprint identified in the fingerprint image to be matched. do,
Fingerprint processing method. to the computer,
Determining a representative minutiae to be used in the fingerprint matching process based on a first distance to a pixel region in the fingerprint image whose image sharpness is unknown among a plurality of minutiae of the fingerprint identified in the fingerprint image to be matched. A program that makes Determining a representative minutiae to be used in the fingerprint matching process based on a first distance to a pixel region in the fingerprint image whose image sharpness is unknown among a plurality of minutiae of the fingerprint identified in the fingerprint image to be matched. do,
Fingerprint processing circuit.

Images