JP6442873B2 - Fingerprint image processing apparatus, fingerprint image processing method and program - Google Patents

Description

  The present invention relates to a fingerprint image processing apparatus, a fingerprint image processing method, and a program.

  Fingerprint images are used for purposes such as immigration, criminal investigation and personal authentication on information terminals. In recent years, a fingerprint collation system using a database in which hundreds of thousands or more of fingerprint images are stored is operating.

  In such a fingerprint collation system, an extraction target image (collation target candidate) is often extracted from a huge number of fingerprint images stored in a database. By narrowing down the fingerprint images to be collated, the processing required for fingerprint collation is reduced and the speed is improved.

  For example, in Patent Document 1, a rough feature shape of a fingerprint ridge is extracted by performing dimension compression processing (KL expansion) on a fingerprint image, and a fingerprint image that matches the feature shape is extracted as a collation target image from a database. A system (image feature extraction apparatus) is disclosed (see FIG. 20).

Japanese Patent Laid-Open No. 10-177650

  The disclosure of the above prior art document is incorporated herein by reference. The following analysis was made by the present inventors.

  The image feature extraction apparatus disclosed in Patent Document 1 is an apparatus that narrows down a collation target by determining a feature amount of each fingerprint image based on the direction of a fingerprint ridge. However, if the collation target is narrowed down based on the feature amount calculated from the direction of the fingerprint ridge, there is a possibility that the collation target candidate may actually be used despite the fingerprint image of a different person.

  For example, the fingerprint images shown in FIGS. 21A and 21B are fingerprint images of different persons. If a person who is not familiar with fingerprint identification compares the two fingerprint images shown in FIGS. 21A and 21B, it is likely to be regarded as a fingerprint image of the same person. Since the directions in which the fingerprint ridges are directed approximately coincide with each other at the corresponding portions of both fingerprint images, both fingerprint images can be recognized as fingerprint images of the same person.

  This is also the case with the image feature extraction device disclosed in Patent Document 1, and there is a possibility that two fingerprint images may be determined as fingerprint images of the same person (extracted as collation target candidates). FIG. 22 is a drawing in which fingerprint ridges are extracted from the fingerprint images of FIGS. 21 (a) and 21 (b). 22 (a) and 22 (b), if the direction of the fingerprint ridge is quantified and the quantified direction of the fingerprint ridge is extracted as a feature amount of each image, the directions of the fingerprint ridge are approximately the same. In the image feature extraction device of Patent Document 1, there is a possibility that one fingerprint image is determined as the other verification target candidate.

  However, not only the direction of the fingerprint ridge but also the position thereof, it becomes clear that both fingerprint images are fingerprint images of different persons. For example, comparing the region 601 in FIG. 22A and the region 602 in FIG. 22B, it can be confirmed that there are fingerprint ridges in different directions in the two regions. Specifically, the direction of the fingerprint ridge in the region 602 in FIG. 22B is different from the direction of the fingerprint ridge included in the region 601 in FIG. There are other places like this, for example, the directions of the fingerprint ridges included in the region 603 and the region 604 are different from each other. That is, not only the direction of the fingerprint ridge but also the position thereof, it is found that the two fingerprint images are fingerprint images of different persons.

  As described above, the image feature extraction device disclosed in Patent Document 1 has room for improving accuracy when extracting a fingerprint image to be collated.

  An object of the present invention is to provide a fingerprint image processing device, a fingerprint image processing method, and a program that contribute to enabling extraction of a fingerprint image to be collated with high accuracy.

  According to the first aspect of the present invention, the feature amount is related to the feature point direction of each of the plurality of feature points in the fingerprint image to be searched, and is calculated from the rotation direction of the feature point direction with respect to the reference point in the fingerprint image. A plurality of second features calculated from the first feature amount vector having the plurality of first feature amounts as elements and the feature amount relating to the feature point direction and the direction of the feature point direction with respect to the reference point A fingerprint image processing apparatus is provided that includes a feature quantity vector calculation unit that calculates a second feature quantity vector having a quantity as an element.

  According to the second aspect of the present invention, the feature amount is related to the feature point direction of each of the plurality of feature points in the fingerprint image to be searched, and is calculated from the rotation direction of the feature point direction with respect to the reference point in the fingerprint image. Calculating a first feature quantity vector having a plurality of first feature quantities as elements, and a plurality of feature quantities related to the feature point direction and calculated from directions of the feature point direction with respect to the reference point A second feature quantity vector having the second feature quantity as an element is calculated.

According to the third aspect of the present invention, the feature amount is related to the feature point direction of each of the plurality of feature points in the fingerprint image to be searched and is calculated from the rotation direction of the feature point direction with respect to the reference point in the fingerprint image. A plurality of processing for calculating a first feature quantity vector having a plurality of first feature quantities as elements, and a feature quantity related to the feature point direction and calculated from an orientation of the feature point direction with respect to the reference point There is provided a program for causing a computer that controls the fingerprint image processing apparatus to execute a process of calculating a second feature quantity vector having the second feature quantity as an element.
This program can be recorded on a computer-readable storage medium. The storage medium may be non-transient such as a semiconductor memory, a hard disk, a magnetic recording medium, an optical recording medium, or the like. The present invention can also be embodied as a computer program product.

  According to each aspect of the present invention, a fingerprint image processing apparatus, a fingerprint image processing method, and a program that contribute to enabling extraction of a fingerprint image to be collated with high accuracy are provided.

It is a figure for demonstrating the outline | summary of one Embodiment. It is a figure which shows an example of the whole structure of the collation target extraction system which concerns on 1st Embodiment. It is a figure which shows an example of an internal structure of a collation target extraction apparatus. It is a figure which shows an example of an internal structure of a collation target determination part. It is a figure for demonstrating the feature point which a feature point extraction part extracts. It is a figure which shows an example of the feature point direction of each feature point which extracted the 1st and 2nd feature point from the fingerprint image. It is a figure for demonstrating a 1st feature-value. It is an example of the 1st feature-value image corresponding to each feature point shown in FIG. It is a figure for demonstrating the 2nd feature-value. It is an example of the 2nd feature-value image corresponding to each feature point shown in FIG. It is a figure for demonstrating the process of a 1st feature-value image smoothing part. It is an example of the image obtained by smoothing the 1st feature-value image shown in FIG. It is an example of the image obtained by smoothing the 2nd feature-value image shown in FIG. It is a figure for demonstrating calculation of a 1st feature-value vector. It is a figure for demonstrating calculation of a 2nd feature-value vector. It is a figure which shows an example of a 1st feature-value image. It is a figure which shows an example of a 2nd feature-value image. 5 is a flowchart illustrating an example of the operation of the verification target extraction device 30. It is a figure which shows another example of the internal structure of a collation target determination part. It is a figure which shows an example of the extraction process of the fingerprint image used as collation object from a database. It is a figure which shows an example of a fingerprint image. It is drawing which extracted the fingerprint ridgeline from the fingerprint image of Fig.21 (a) and FIG.21 (b).

  First, an outline of an embodiment will be described with reference to FIG. Note that the reference numerals of the drawings attached to the outline are attached to the respective elements for convenience as an example for facilitating understanding, and the description of the outline is not intended to be any limitation.

  As described above, a fingerprint image processing apparatus that can extract a fingerprint image to be collated with high accuracy is desired.

  Therefore, as an example, the fingerprint image processing apparatus 100 shown in FIG. 1 is provided. The fingerprint image processing apparatus 100 includes a feature quantity vector calculation unit 101. The feature amount vector calculation unit 101 is a feature amount related to the feature point direction of each of the plurality of feature points in the fingerprint image to be searched, and is calculated from the rotation direction of the feature point direction with respect to the reference point in the fingerprint image. A second feature quantity having a plurality of second feature quantities calculated from a first feature quantity vector having the first feature quantity as an element and a feature quantity relating to the feature point direction and the direction of the feature point direction with respect to the reference point. A feature vector is calculated.

  The feature quantity vector calculated by the fingerprint image processing apparatus 100 is obtained by quantifying the feature point directions at a plurality of feature points dispersed in the fingerprint image. Therefore, a fingerprint image (a fingerprint vector of the same person is searched from the feature vector (feature vector on the matching target side) calculated from a large number of fingerprint images stored in the database and the fingerprint image stored in the database ( By comparing the feature quantity vector calculated from the search target fingerprint image) (the feature quantity vector on the search target side), the fingerprint image to be collated can be extracted with high accuracy. The reason is that when calculating the feature vector on the search target side and the feature vector on the collation target side, the feature points at the same coordinate position are selected, so that two feature amounts considering the distribution of the feature points in the fingerprint image are taken into account. This is because a vector is obtained. For example, by calculating the similarity between two feature vectors and performing threshold processing on the calculated similarity, a fingerprint image to be collated (a fingerprint image stored in a database) can be a candidate for collation. Can be judged.

  In addition, although the feature of the fingerprint image indicated by the feature point itself is merely position information, the fingerprint image processing apparatus 100 calculates another feature amount called the feature point direction from the feature point, and in the subsequent processing, It can be used for extraction. Even if a feature point exists in the same position in the fingerprint image on the search side and the collation side, it is rare to match up to the feature amount (rotation direction and direction with respect to the reference point) regarding the feature point direction. If the feature quantity vector calculated by the fingerprint image processing apparatus 100 is used, it is possible to determine the match / mismatch more accurately for such a fingerprint image. That is, the fingerprint image processing apparatus 100 can extract a finger image to be collated with high accuracy.

  In this document, the feature point direction of a feature point is defined as a direction in which a ridge is generated from the feature point or a direction in which a ridge that branches at the feature point proceeds.

  Hereinafter, specific embodiments will be described in more detail with reference to the drawings.

[First Embodiment]
The first embodiment will be described in more detail with reference to the drawings.

  FIG. 2 is a diagram illustrating an example of the overall configuration of the verification target extraction system according to the first embodiment. Referring to FIG. 2, the verification target extraction system includes a fingerprint imaging device 10, a database 20, and a verification target extraction device 30. The function of the above-described fingerprint image processing apparatus 100 is included in the verification target extraction apparatus 30.

  The fingerprint imaging device 10 and the verification target extraction device 30 are connected by communication means such as USB (Universal Serial Bus). The database 20 and the verification target extraction device 30 are configured to be able to communicate with each other via a network.

  The fingerprint photographing device 10 is a means for photographing a human fingerprint and acquiring a fingerprint image. The fingerprint imaging device 10 is a so-called scanner, and captures a fingerprint and outputs a color, gray scale, or monochrome fingerprint image to the collation target extraction device 30. Hereinafter, the fingerprint image output by the fingerprint imaging apparatus 10 is referred to as an input fingerprint image.

  The database 20 is a means for storing a plurality of fingerprint images to be collated with the input fingerprint image. The database 20 outputs a fingerprint image in response to a request from the verification target extraction device 30. Hereinafter, the fingerprint image output from the database 20 is referred to as a database fingerprint image.

  In response to the input fingerprint image, the verification target extraction device 30 selects a fingerprint image that is considered to match the input fingerprint image from among a plurality of database fingerprint images stored in the database 20, and outputs the fingerprint image as a verification target candidate. Means.

  The verification target candidates are used in a verification device (not shown) connected to the verification target extraction device 30. The collation device is a device that identifies a fingerprint image (database fingerprint image) that is considered to have the highest probability of matching an input fingerprint image from among a plurality of collation target candidates.

  FIG. 3 is a diagram illustrating an example of an internal configuration of the verification target extraction device 30. Referring to FIG. 3, the verification target extraction device 30 includes a communication unit 201, an input unit 202, a display unit 203, an output unit 204, a storage unit 205, and a verification target determination unit 206. The

  The communication unit 201 is a means for communicating with the database 20 and acquiring a database fingerprint image.

  The input unit 202 includes a driver for the fingerprint imaging apparatus 10 and is a unit that acquires an input fingerprint image.

  The display unit 203 is a means for notifying the operation state of the verification target extraction device 30 to the operator.

  The output unit 204 is a means for outputting the collation target candidate output from the collation target determination unit 206 to the outside.

  The storage unit 205 is a unit that stores programs, data, and the like necessary for the operation of the verification target extraction device 30.

  The collation target determining unit 206 includes a search-side feature vector (hereinafter referred to as feature vector Vi) corresponding to the input fingerprint image and a collation-side feature vector (hereinafter referred to as feature vector Vd) corresponding to the database fingerprint image. The database fingerprint image that can be a candidate for collation is extracted (selected) by comparing these feature quantity vectors. The collation target determination unit 206 delivers the database fingerprint image determined to be a collation target candidate to the output unit 204 as a collation target candidate. The feature amount vectors Vi and Vd corresponding to the two fingerprint images will be described later.

  Moreover, the structure of the collation target extraction device 30 shown in FIG. 3 is an exemplification, and various modifications can be considered. For example, the output unit 204 may be deleted, and the verification target candidates may be displayed on the display unit 203. Or you may output a collation target candidate to the terminal (for example, personal computer, a smart phone, etc.) network-connected with the collation target extraction apparatus 30 using the communication part 201. FIG.

  Next, the verification target determination unit 206 will be described.

  FIG. 4 is a diagram illustrating an example of an internal configuration of the collation target determination unit 206. Referring to FIG. 4, the collation target determination unit 206 includes a feature quantity vector calculation unit 301 and a collation target candidate extraction unit 302.

  Further, the feature quantity vector calculation unit 301 includes a feature point extraction unit 401, a first feature quantity image generation unit 402, a second feature quantity image generation unit 403, a first feature quantity image smoothing unit 404, and a second feature. An amount image smoothing unit 405, a first feature amount vector calculation unit 406, a second feature amount vector calculation unit 407, and a feature amount vector generation unit 408 are configured.

<Feature point extraction>
The feature point extraction unit 401 is a means for extracting feature points included in a fingerprint image (database fingerprint image, input fingerprint image) input to the verification target extraction device 30.

  The feature point extraction unit 401 extracts two types of feature points from the fingerprint image. FIG. 5 is a diagram for explaining the feature points extracted by the feature point extraction unit 401.

  The first feature point targeted by the feature point extraction unit 401 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 feature point extraction unit 401 extracts a ridge end point such as the first feature point P1 shown in FIG. 5A as the first feature point of the input fingerprint image. Specifically, the feature point extraction unit 401 extracts the first feature points while scanning the entire fingerprint image.

  The second feature point targeted by the feature point extraction unit 401 is a ridge branch point. FIG. 5B illustrates a second feature point P2 where the ridge 511 branches into two ridges 512, 513. The feature point extraction unit 401 extracts a ridge branch point such as the second feature point P2 shown in FIG. 5B as the second feature point of the input fingerprint image.

  Next, the feature point extraction unit 401 calculates the feature point direction of each extracted feature point. As for the first feature point, as shown in FIG. 5A, the direction in which the ridge 503 is generated from the first feature point P1 (the direction of the arrow 504) is defined as the feature point direction of the first feature point. .

  As for the second feature point, as shown in FIG. 5B, when the ridge 511 is branched into two ridges 512, 513 at the second feature point P2, the second feature point P2 to 2 The direction (in the direction of the arrow 514) traveling between the ridges of the book (in the middle of the vector) is defined as the feature point direction of the second feature point.

  The extracted position information (coordinates) related to the first feature point and the second feature point is stored in the form of pixel positions in the fingerprint image. Further, the angle between the X axis and the feature point direction line in the XY coordinate system in which each feature point is set as the origin is stored as the feature point direction of each feature point.

  FIG. 6 is a diagram illustrating an example of the feature point direction of each extracted feature point by extracting the first and second feature points from the fingerprint image. In FIG. 6, circles indicate the first feature points (end points) described above, and squares indicate the second feature points (branch points). A straight line generated from each feature point is a feature point direction line indicating the feature point direction of each feature point.

  In the first embodiment, the feature point extraction unit 401 is described as extracting two types of feature points, but the type of feature points to be extracted is not limited. The feature point extraction unit 401 may extract at least one of the first feature point (end point) and the second feature point (branch point).

<Generation of First Feature Quantity Image>
The first feature value image generation unit 402 generates a first feature value image based on each feature point extracted by the feature point extraction unit 401 and the feature point direction of each feature point.

  The first feature amount image is an image having a pixel value of a pixel forming a fingerprint image as a first feature amount. The first feature amount is a direction in which the feature point direction of each feature point with respect to a first reference line connecting each feature point and the center point of the fingerprint image (also referred to as a reference point or a core of the fingerprint image) is directed (rightward direction). , Leftward) is a quantified feature amount. Alternatively, the first feature amount can be said to be a feature amount obtained by quantifying the rotation direction (clockwise or counterclockwise) of the feature point direction with respect to the center point of the fingerprint image.

  FIG. 7 is a diagram for explaining the first feature amount. FIG. 7 illustrates the center point (core) of the fingerprint image and the first feature point P1.

  The first feature image generation unit 402 calculates a straight line connecting the center point of the fingerprint image and the feature point as a first reference line.

  Next, the first feature quantity image generation unit 402 calculates an angle θ formed by the first reference line and the feature point direction line. Specifically, the first feature quantity image generation unit 402 sets the origin of the XY coordinate axes to the first feature point P1, and calculates an angle α1 formed by the X axis and the first reference line (FIG. 7A). reference). Next, the first feature quantity image generation unit 402 reads the feature point direction line angle α2 stored when the first feature point P1 is extracted. Next, the first feature image generation unit 402 subtracts the angle α2 from the angle α1 to calculate the angle θ. The angles α1 and α2 are positive values when they are in the first and second quadrants of the XY coordinate system, and negative values when they are in the third and fourth quadrants.

  As a result of the calculation, if the angle θ is a positive value, the feature point direction of each feature point is determined to be rightward with respect to the first reference line (clockwise with respect to the center point). On the other hand, if the angle θ is a negative value as a result of the calculation, it is determined that the feature point direction of each feature point is leftward with respect to the first reference line (counterclockwise with respect to the center point).

  When it is determined that the feature point direction is clockwise (clockwise) with respect to the center point from the calculated angle θ, the first feature quantity image generation unit 402 moves in the feature point direction of the feature point. On the other hand, a first feature amount in the range of 128 to 255 (hexadecimal display; 80 to FF) is assigned.

  More specifically, as shown in FIG. 7B, when the angle θ1 is 0 degree (when the first reference line and the feature point direction are in the same direction), 128 is given to the first feature amount. It is done. As the angle θ1 increases, the first feature amount also increases. When the angle θ1 is 90 degrees, 255, which is the maximum value of the first feature amount, is given. Further, when the angle θ1 increases beyond 90 degrees, the first feature amount decreases, and when the angle θ1 is 180 degrees (when the first reference line and the feature point direction are opposite), the first feature amount is decreased. 128 is given as the feature amount.

  As described above, when the feature point direction line of the feature point is directed to the right with respect to the first reference line, the first feature amount image generation unit 402 has the first feature in the range of 128 to 255 with respect to the feature point direction. Give the amount.

  On the other hand, if it is determined from the calculated angle θ that the feature point direction is counterclockwise (counterclockwise) with respect to the center point, the first feature amount image generation unit 402 determines the feature point feature. A first feature amount in the range of 0 to 128 (hexadecimal display; 00 to 80) is assigned to the point direction.

  More specifically, as shown in FIG. 7C, when the angle θ2 is 0 degree (when the first reference line and the feature point direction are in the same direction), 128 is given to the first feature amount. It is done. As the angle θ2 increases, the first feature value decreases. When the angle θ2 is 90 degrees, 0, which is the minimum value of the first feature value, is given. Further, when the angle θ2 increases beyond 90 degrees, the first feature amount increases. When the angle θ2 is 180 degrees (when the first reference point and the feature point direction are opposite), the first feature amount is increased. 128 is given as the feature amount.

  As described above, when the feature point direction line of the feature point is facing left with respect to the first reference line, the first feature amount image generation unit 402 has a first feature in the range of 0 to 128 with respect to the feature point direction. Give the amount.

  When the first reference line and the feature point direction line are in the same direction (or in the opposite direction), either the process described using the angle θ1 or the process described using the angle θ2 may be performed. Is natural.

  The first feature value image generation unit 402 performs the first feature value calculation process for each feature point extracted by the feature point extraction unit 401.

  Next, the first feature value image generation unit 402 generates a first feature value image based on the calculated first feature value.

  Specifically, the first feature quantity image generation unit 402 regards each feature point as the origin in the XY coordinate system, and sets the pixel value of pixels included in a predetermined range from the origin (feature point) as the first feature quantity. change. The predetermined range from the origin is a predetermined range such as several pixels in the X direction and several pixels in the Y direction from the origin. In addition, an area where the periphery (predetermined range) of each feature point overlaps is an average value of the first feature amount for each overlapping area. For example, if the first feature amounts corresponding to two overlapping regions are 0 and 255, the first feature amount of the overlapping region is 127. The first feature amount image generation unit 402 assigns 128 as the pixel value (first feature amount) to pixels that are not around the feature points.

  In this way, the first feature value image generation unit 402 generates a first feature value image. FIG. 8 is an example of a first feature amount image corresponding to each feature point shown in FIG. In FIG. 8, the first feature value is expressed in correspondence with the density so that the first feature value becomes visual. Specifically, black is assigned to the minimum value 0 of the first feature value, gray is assigned to the intermediate value 128, and white is assigned to the maximum value 255.

<Generation of Second Feature Quantity Image>
The second feature amount image generation unit 403 generates a second feature amount image based on each feature point extracted by the feature point extraction unit 401 and the feature point direction of each feature point.

  The second feature amount image is an image having the pixel value of the pixels forming the fingerprint image as the second feature amount. The second feature amount is a feature amount obtained by quantifying the directionality with respect to the center point of the feature point direction line.

  FIG. 9 is a diagram for explaining the second feature amount. The second feature amount image generation unit 403 calculates a straight line connecting the center point of the fingerprint image and the feature point as the first reference line. Further, the second feature quantity image generation unit 403 calculates a second reference line orthogonal to the first reference line with the feature point as an intersection.

  Next, the second feature amount image generation unit 403 calculates an angle Φ formed by the second reference line and the feature point direction line. Specifically, the second feature quantity image generation unit 403 sets the origin of the XY coordinate axis to the first feature point P1, and calculates the angle β1 formed by the Y axis and the second reference line (FIG. 9A). reference). Next, the second feature quantity image generation unit 403 reads the angle α2 of the feature point direction line stored when extracting the first feature point P1, and the angle β2 formed between the feature point direction line and the Y axis. Calculate Next, the second feature quantity image generation unit 403 calculates the angle Φ by subtracting the angle β2 from the angle β1. The angles β1 and β2 are positive values when they are in the second and third quadrants of the XY coordinate system, and negative values when they are in the first and fourth quadrants.

  If the calculated angle Φ is a positive value, the feature point direction of each feature point is determined to be a direction away from the center point (withdrawal direction). On the other hand, if the calculated angle Φ is a negative value, the feature point direction of each feature point is determined to be a direction approaching the center point (approach direction).

  When it is determined from the calculated angle Φ that the feature point direction is a direction away from the center point, the second feature amount image generation unit 403 has 0 to 128 with respect to the feature point direction of the feature point. A second feature quantity in the range is assigned.

  More specifically, as shown in FIG. 9B, when the angle Φ1 is 0 degree (when the second reference line and the feature point direction are in the same direction), 128 is given to the second feature amount. It is done. As the angle Φ1 increases, the second feature value decreases. When the angle Φ1 is 90 degrees, 0, which is the minimum value of the second feature value, is given. Further, when the angle Φ1 increases beyond 90 degrees, the second feature amount increases. When the angle Φ1 is 180 degrees (when the second reference line and the feature point direction are opposite), the second feature amount is increased. 128 is given as the feature amount.

  As described above, when the feature point direction line of the feature point is away from the center point, the second feature amount image generation unit 403 gives the second feature amount in the range of 0 to 128 with respect to the feature point direction. .

  On the other hand, when it is determined from the calculated angle Φ that the feature point direction is a direction approaching the center point, the second feature quantity image generation unit 403 performs 128 to 128 with respect to the feature point direction of the feature point. A second feature amount in the range of 255 is assigned.

  More specifically, as shown in FIG. 9C, when the angle Φ2 is 0 degree (when the second reference line and the feature point direction are in the same direction), 128 is given to the second feature amount. It is done. As the angle Φ2 increases, the second feature amount also increases. When the angle Φ2 is 90 degrees, 255, which is the maximum value of the second feature amount, is given. Further, when the angle Φ2 increases beyond 90 degrees, the second feature amount decreases, and when the angle Φ2 is 180 degrees (when the second reference line and the feature point direction are opposite), the second feature amount is decreased. 128 is given as the feature amount.

  As described above, when the feature point direction line of the feature point approaches the center point, the second feature amount image generation unit 403 gives the second feature amount in the range of 128 to 255 with respect to the feature point direction. .

  The second feature amount image generation unit 403 performs the second feature amount calculation process for each feature point extracted by the feature point extraction unit 401.

  Next, the second feature amount image generation unit 403 generates a second feature amount image based on the calculated second feature amount. Since the second feature value image has no difference from the first feature value image except that the feature value used for the calculation is the second feature value, description thereof is omitted.

  FIG. 10 is an example of a second feature amount image corresponding to each feature point shown in FIG. Referring to FIG. 10, since there are many feature points in the direction in which the feature point direction moves away from the center point in the lower region, it can be seen that there are many regions with colors close to black. On the other hand, since the feature point direction in the upper region is neither a direction away from the center point nor a direction approaching it, it can be seen that there are many regions of colors close to gray.

  Various methods are conceivable for calculating the center point of the fingerprint image used when generating the first and second feature amount images. For example, a portion having the highest ridge peak included in the fingerprint image can be calculated as the center point of the fingerprint image. However, it is not intended to limit the calculation of the center point.

<Smoothing process>
Although the details will be described later, the first and second feature quantity images are used to calculate the feature quantity vectors of the database fingerprint image and the input fingerprint image, and the database fingerprint image is obtained by comparing the two feature quantity vectors. It is determined whether the image can be a candidate for collation. At that time, since the database fingerprint image and the input fingerprint image are usually different in terms of conditions such as a fingerprint photographing apparatus and method, there are often some differences even in the fingerprint image of the same person. . For example, there may be a slight difference in the position of each feature point even in the fingerprint image of the same person due to the tilt of the finger at the time of fingerprint shooting. In order to absorb such errors and the like, the first and second feature amount images are smoothed. Hereinafter, the smoothing process of the first and second feature amount images will be described.

  The first feature value image smoothing unit 404 performs a process of smoothing the first feature value image generated by the first feature value image generating unit 402.

  Specifically, the first feature amount image smoothing unit 404 calculates the average value of the first feature amounts of the reference pixel and its surrounding pixels, using each pixel position of the first feature amount image as a reference. The first feature amount of the pixel. Note that a range used for smoothing the first feature amount image is defined as a smoothing range.

  FIG. 11 is a diagram for explaining the processing of the first feature quantity image smoothing unit 404. In FIG. 11, black circles, white circles, and gray circles mean pixels that form the first feature amount image. Further, in FIG. 11, it is assumed that a range of one pixel from each pixel is set as a smoothing range.

  In this case, for example, when calculating the first feature amount of the pixel 520, the first feature amount of the pixel 520 and the first feature amount of each of the surrounding pixels 521 to 528 are added, and the added pixel The first feature amount of the pixel 520 is obtained by dividing by the number of pixels (9 in the case of FIG. 11). The first feature amount image smoothing unit 404 smoothes the first feature amount by applying such smoothing processing (average value calculation processing) to each pixel constituting the first feature amount image.

  FIG. 12 is an example of an image obtained by smoothing the first feature amount image shown in FIG. The smoothing range may be a previously fixed range or a variable range. For example, the smoothing range may be changed according to the location of the first feature amount image. Specifically, the smoothing range may be made finer around the center point, and the smoothing range may be made coarser as the distance from the center point increases.

  The second feature amount image smoothing unit 405 smoothes the second feature amount image. FIG. 13 is an example of an image obtained by smoothing the second feature amount image shown in FIG. Since the process performed by the second feature amount image smoothing unit 405 is not different from the process performed by the first feature amount image smoothing unit 404, the description thereof is omitted.

<Calculation of feature vector>
The first feature quantity vector calculation unit 406 calculates a first feature quantity vector from the smoothed first feature quantity image.

  The first feature quantity vector calculation unit 406 extracts a plurality of pixels from the smoothed first feature quantity image, and calculates a vector in which the extracted first feature quantities of each pixel are arranged as a first feature quantity vector. . Specifically, the first feature vector calculation unit 406 sets grid-like (matrix) sampling coordinates on the smoothed first feature image. The first feature quantity vector calculation unit 406 calculates a first feature quantity vector by arranging the first feature quantities of pixels corresponding to the respective sampling coordinates.

For example, as shown in FIG. 14, grid-like sampling coordinates are set on the first feature value image so as to form a matrix of m rows and n columns (n and m are both positive integers; the same applies hereinafter). The first feature quantity vector calculation unit 406 calculates the first feature quantity vector v1 that can be expressed by Expression (1) by reading the first feature quantity of each sampling coordinate.

v1 = (FV1_11, FV1_12,..., FV1_n1,... FV1_nm) (1)

FV1 indicates a first feature amount (FV1; Feature Value 1), and sampling coordinates are indicated by numbers following the underbar. For example, FV1_11 indicates the first feature amount of the sampling coordinates (1, 1) in FIG.

  Similarly, the second feature quantity vector calculation unit 407 also sets grid-like sampling coordinates (see FIG. 15) in the smoothed second feature quantity image, and arranges the second feature quantities corresponding to the sampling coordinates, A second feature vector is calculated. Note that the processing performed by the second feature quantity vector calculation unit 407 is not different from the processing performed by the first feature quantity vector calculation unit 406, and thus description thereof is omitted.

The second feature quantity vector calculation unit 407 performs the same processing as the first feature quantity vector calculation unit 406, thereby calculating the second feature quantity vector v2 that can be expressed by Expression (2).

v2 = (FV2_11, FV2_12,..., FV2_n1,... FV2_nm) (2)

FV2 indicates the second feature amount, and the sampling coordinates are indicated by numbers following the underbar.

The feature vector generation unit 408 combines the first feature vector v1 and the second feature vector v2 to output a feature vector V corresponding to the input fingerprint image. Specifically, the feature quantity vector generation unit 408 adds the second feature quantity vector v2 represented by the expression (2) to the first feature quantity vector v1 represented by the expression (1), so that the expression (3 ) To calculate a feature vector V that can be expressed by

V = (FV1_11,..., FV1_nm, FV2_11,..., FV2_nm) (3)

  The feature vector generation unit 408 outputs the combined first and second feature vectors to the collation target candidate extraction unit 302.

  The feature vector calculation unit 301 calculates the feature vectors Vd and Vi for each fingerprint image (database fingerprint image and input fingerprint image) to be input, and outputs the calculated feature vectors Vd and Vi to the collation target candidate extraction unit 302. Therefore, it can be said that the feature vector Vd and the feature vector Vi are feature vectors composed of the same kind of elements.

<Determination of candidates for verification>
The collation target candidate extraction unit 302 determines whether or not the database fingerprint image corresponding to the input feature vector Vd is sufficient to collate with the input fingerprint image corresponding to the feature vector Vi.

  When it is determined that the database fingerprint image can be a collation target, the collation target candidate extraction unit 302 extracts the database fingerprint image as a collation target candidate and outputs it. On the other hand, when it is determined that the database fingerprint image cannot be a collation target, the collation target candidate extraction unit 302 discards the database fingerprint image. When the database fingerprint image is discarded, the collation target extraction device 30 acquires the next database fingerprint image from the database 20 and repeats the collation target extraction process.

  Next, the collation target narrowing process of the collation target candidate extraction unit 302 will be described.

  The verification target candidate extraction unit 302 calculates the difference between the feature quantity vector Vd and the feature quantity vector Vi. Specifically, the collation target candidate extraction unit 302 calculates a difference value between corresponding elements in the feature quantity vector Vd and the feature quantity vector Vi. Thereafter, the collation target candidate extraction unit 302 calculates the total difference value of the calculated elements.

  Next, when the sum of difference values is equal to or less than a predetermined threshold, the collation target candidate extraction unit 302 extracts a database fingerprint image corresponding to the input feature vector Vd as a collation target, and outputs it as a collation target candidate. To do.

  On the other hand, the collation target candidate extraction unit 302 discards the database fingerprint image when the sum of the difference values is larger than a predetermined threshold.

  FIG. 16 is a diagram illustrating an example of the first feature amount image. FIG. 16A is an example of an input fingerprint image and its first feature value image. FIG. 16B is an example of a database fingerprint image of the same person as the input fingerprint image of FIG. 16A and a first feature amount image thereof. FIG. 16C is an example of a database fingerprint image of a person different from the input fingerprint image of FIG.

  Referring to the fingerprint images in FIGS. 16A to 16C, they all seem to be fingerprint images of the same person.

  Comparing the three first feature amount images, it can be understood that the distributions of the black and white regions are similar in FIGS. 16 (a) and 16 (b). That is, the distribution of the first feature amount is similar between FIG. 16 (a) and FIG. 16 (b). On the other hand, it can be understood that the first feature quantity image of FIG. 16C has a clear black area in the area above the first feature quantity image of FIG. That is, the distribution of the first feature amount is not similar between FIG. 16 (a) and FIG. 16 (c).

  Such a difference or non-difference in the distribution of the first feature value is reflected in the difference value between the first feature value vectors. For example, in FIG. 16A and FIG. 16B, if the sampling coordinates are the same, the first feature amount at the sampling coordinates is considered to be almost the same. That is, between the elements of the first feature quantity vector calculated from the first feature quantity image shown in FIG. 16A and the first feature quantity vector calculated from the first feature quantity image shown in FIG. The sum of the difference values is a small value.

  On the other hand, in FIG. 16A and FIG. 16C, even if the sampling coordinates are the same, the first feature amount at the sampling coordinates is greatly different. Therefore, between the elements of the first feature quantity vector calculated from the first feature quantity image shown in FIG. 16A and the first feature quantity vector calculated from the first feature quantity image shown in FIG. The sum of the difference values is a large value.

  As described above, since the difference in the distribution of the first feature values is reflected in the difference value between the first feature value vectors, it is possible to narrow down the database fingerprint images that are candidates for matching by appropriately determining the threshold value. Specifically, the threshold value is set low when it is desired to increase the accuracy of narrowing down, and the threshold value is set high when a large number of collation target candidates are desired.

  As described above, the comparison target candidate can be extracted by comparing the first feature vector. However, the first feature amount only reflects the feature regarding the rotation direction (clockwise, counterclockwise) of the feature point direction of each feature point. Therefore, for example, referring to FIG. 7, the first feature amount to be given is the same regardless of whether the feature point direction line is away from or close to the center point.

  However, since the orientation with respect to the center point is considered to be a valuable feature in the fingerprint image, the candidate for collation can be extracted with higher accuracy by considering the feature. Therefore, the collation target extraction device 30 uses such a feature amount (second feature amount; the direction of the feature point direction with respect to the center point) for determination of a collation target candidate. Specifically, the collation target extraction device 30 uses the second feature quantity vector calculated from the second feature quantity image to determine a collation target candidate.

  FIG. 17 is a diagram illustrating an example of the second feature amount image. FIG. 17 shows an example of a second feature amount image corresponding to each fingerprint image shown in FIG. Referring to FIG. 17, it can be understood that the difference in the second feature amount appears in the lower region of the fingerprint image. This is probably because the ridges in the upper region of many human fingerprint images are often concentric. On the other hand, the ridges in the lower region of the fingerprint image are not concentric, and the pattern is greatly different among individuals.

  The second feature vector mainly reflects the difference in the lower area of the fingerprint image. Therefore, by applying a discrimination process using a threshold value in the same manner as the first feature vector, a more accurate collation target candidate can be determined.

  The operation of the collation target extraction device 30 is summarized as shown in the flowchart of FIG.

  First, the collation target determining unit 206 inputs a fingerprint image (database fingerprint image, input fingerprint image) (step S101). Next, the collation target determination unit 206 extracts feature points from the fingerprint image and calculates the feature point direction for each feature point (step S102).

  The collation target determining unit 206 generates a first feature amount image having the first feature amount relating to each extracted feature point as a pixel value (step S103). Next, the collation target determination unit 206 smoothes the generated first feature value image (step S104). Thereafter, the collation target determining unit 206 calculates a first feature vector corresponding to the fingerprint image from the smoothed first feature image (step S105).

  The collation target determining unit 206 generates a second feature amount image having the second feature amount relating to each extracted feature point as a pixel value (step S106). Next, the collation target determining unit 206 smoothes the generated second feature amount image (step S107). Thereafter, the collation target determining unit 206 calculates a second feature quantity vector corresponding to the fingerprint image from the smoothed second feature quantity image (step S108).

  When the calculation of the first and second feature quantity vectors for each of the database fingerprint image and the input fingerprint image is completed, the collation target determining unit 206 is based on the feature quantity vector Vd and the feature quantity vector Vi corresponding to each fingerprint image. The candidate for collation is extracted (step S109).

  Note that the flowchart shown in FIG. 18 is an example, and is not intended to limit the execution order of each step. For example, the second feature quantity vector may be calculated before the first feature quantity vector, or two feature quantity vectors may be calculated in parallel.

  Although the verification target extraction system has been described with reference to FIG. 2, the system configuration is not limited to the configuration of FIG. For example, a storage unit corresponding to the database 20 may be included in the verification target extraction device 30, or the verification target extraction device 30 may include a scanner function for photographing a fingerprint. Alternatively, the input fingerprint image may be transmitted via a network and input to the verification target extraction device 30.

  Although the database 20 has been described as being stored with fingerprint images, the database 20 may store feature vectors of accumulated fingerprint images. In this case, instead of acquiring the database fingerprint image, the verification target extraction device 30 acquires a feature amount vector, and compares the feature amount vector with a feature amount vector calculated from the input fingerprint image. By adopting such a configuration, it is possible to execute a process related to determination of a candidate for collation at high speed.

  Alternatively, every time the collation target extraction device 30 calculates a feature vector of a database fingerprint image, the feature vector is written back to the database 20, and the rewritten feature vector is reused as necessary. Also good.

  In the first embodiment, the means for calculating a feature vector from a fingerprint image (feature vector calculation unit 301), the unit for extracting a candidate for collation (collation target candidate extraction unit 302), and collation of the same device The case of being included in the target extraction device 30 has been described. However, these two means may be realized by separate devices.

  In the first embodiment, the feature vector calculating unit 301 is configured to calculate the feature vector Vd (matching-side feature vector) of the database fingerprint image and the feature vector Vi (search-side feature vector) of the input fingerprint image. Explained. However, means for generating these feature quantity vectors may be provided separately. In this case, as shown in FIG. 19, the collation target determination unit 206 is provided with two feature quantity vector calculation units 301a and 301b, and the collation side feature quantity vector and the search side feature quantity vector are collated target candidate extraction unit 302. To provide.

  Alternatively, the functions of the first feature value image smoothing unit 404 and the second feature value image smoothing unit 405, the first feature value vector calculating unit 406, the second feature value vector calculating unit 407, and the like are integrated in one module. It may be realized.

  In the description of the first embodiment, the resolution (number of pixels) of the fingerprint image is not mentioned, but the resolution of the fingerprint image may be changed as necessary. For example, when the resolution of an image obtained from a device that acquires a database fingerprint image and a device that acquires an input fingerprint image (fingerprint imaging device 10) is different, conversion processing such as matching to any resolution may be performed. . Alternatively, if the resolution of the fingerprint image is too high for the use related to extraction of the candidate for collation, the first and second feature amount images may be generated after performing the process of reducing the resolution of the fingerprint image. Good.

  In the first embodiment, when calculating feature amounts (first feature amount, second feature amount), a relative angle of a feature point direction line with respect to a reference line (first reference line, second reference line) is calculated. use. This is because the influence of the distortion of the fingerprint image can be canceled by using the relative angle. For example, when the database fingerprint image is distorted in the right direction and the input fingerprint images acquired from the same person are distorted in the left direction, the absolute angles (feature amounts) in the feature point directions in these fingerprint images are different. Value. However, even if the fingerprint image is distorted, the influence of the distortion is canceled by using a relative angle for calculating the feature amount. In other words, if the database fingerprint image and the input fingerprint image are so good that distortion does not need to be considered, the absolute angle in the XY coordinate system set for the fingerprint image (for example, FIG. 7A) The feature amount may be calculated using the angle α2 of FIG. 9 or the angle β2 of FIG.

  Note that the first feature quantity image smoothing unit 404 and the second feature quantity image smoothing unit 405 perform smoothing of the feature quantity image, but smoothing processing may not be necessary. For example, if both the database fingerprint image and the input fingerprint image have good image quality, the smoothing process is unnecessary.

  Furthermore, in the first embodiment, it is assumed that the amount of information included in each of the database fingerprint image and the input fingerprint image is substantially the same, but in reality, an area in one image is the other image. May not exist. In such a case, even if both fingerprint images are collected from the fingerprints of the same person, when the feature amount vectors of both fingerprint images are compared, the sum of the difference values may be larger than the threshold value. is there. Therefore, when extracting feature points, it is determined whether a part of the fingerprint is missing or the image is not clear, and the feature vector is not compared at a location where an image defect is detected (difference value is not calculated). It is desirable to take measures such as not calculating.

  Alternatively, the quality of each feature quantity may be taken into consideration when comparing the feature quantity vector Vd of the database fingerprint image and the feature quantity vector Vi of the input fingerprint image. Specifically, when extracting feature points from a fingerprint image, the quality of the feature points is also calculated. Then, when the difference value between the elements of the feature quantity vector Vd and the feature quantity vector Vi is calculated, the quality of the detected feature point is reflected. For example, when calculating the difference value between elements of both feature quantity vectors, the smaller of the quality value of the feature point of the database fingerprint image and the quality value of the feature point of the input fingerprint image is used as the difference value. Multiply. When the quality of the feature point in either one of the fingerprint images is extremely low, the quality value indicating the quality is also a very small value. By multiplying such a value by the difference value, a low-quality feature point can be substantially invalidated. Various methods can be used to calculate the quality of feature points.

  The verification target candidate determination method described in the first embodiment can also be used in combination with other verification target candidate determination methods. For example, after the collation target candidates are narrowed down from the shape of the fingerprint ridge disclosed in Patent Document 1, the above-described collation target candidate determination process may be executed.

  The collation target determination unit 206 can also be realized by a computer program that causes a computer installed in the collation target extraction device 30 to execute the processing of each unit in the collation target determination unit 206 using its hardware. Further, by installing the verification target extraction program in the storage unit of the computer, the computer can function as a verification target extraction device. Alternatively, the computer can execute the verification target extraction method by causing the computer to execute the verification target extraction program.

  As described above, the collation target extraction device 30 according to the first embodiment calculates feature quantity vectors for each of the database fingerprint image and the input fingerprint image, and compares these feature quantity vectors to obtain collation target candidates. decide. The database fingerprint image is considered to include a database fingerprint image that is judged to be a fingerprint image of the same person at first glance because the directions of the fingerprint ridges are almost the same. Here, the feature quantity vector calculated in the first embodiment includes not only the feature quantity feature quantity (first feature quantity and second feature quantity) but also information related to the feature quantity distribution. However, even in the case of a database fingerprint image in which the directions of the fingerprint ridges substantially match as described above, since the images that match the distribution of the feature amount are rare, the collation target extraction device according to the first embodiment In 30, the database fingerprint image as described above is not determined as a collation target candidate. That is, it is possible to provide a collation target extraction device that improves the accuracy in selecting a collation target candidate.

A part or all of the above embodiments can be described as in the following supplementary notes, but is not limited thereto.
[Appendix 1]
A plurality of first feature amounts calculated from a rotation direction of the feature point direction with respect to a reference point in the fingerprint image, which is a feature amount of each of the plurality of feature points in the fingerprint image to be searched. A first feature vector,
A feature quantity vector calculation unit for calculating a feature quantity related to the feature point direction and a second feature quantity vector having a plurality of second feature quantities as elements, which is calculated from an orientation of the feature point direction with respect to the reference point A fingerprint image processing apparatus.
[Appendix 2]
A search-side feature quantity vector corresponding to the search target fingerprint image generated from the first and second feature quantity vectors, and a feature quantity vector corresponding to the fingerprint image to be collated, the first and second features A collation target for extracting a collation target candidate for the fingerprint image to be searched from among a plurality of fingerprint images to be collated by comparing the elements of the quantity vector with a collation-side feature vector comprising the same type of elements The fingerprint image processing apparatus according to appendix 1, further comprising a candidate extraction unit.
[Appendix 3]
The fingerprint according to appendix 1 or 2, wherein the feature quantity vector calculation unit calculates the first feature quantity by quantifying an angle formed by a first reference line connecting the reference point and the feature point and a straight line in the feature point direction. Image processing device.
[Appendix 4]
The fingerprint image processing apparatus according to appendix 3, wherein the feature quantity vector calculation unit calculates the second feature quantity by quantifying an angle formed by a second reference line orthogonal to the first reference line and a straight line in the feature point direction. .
[Appendix 5]
The feature vector calculation unit includes:
A feature point extraction unit for extracting the plurality of feature points from the fingerprint image;
A first feature amount image generating unit that generates a first feature amount image by setting a pixel value of the fingerprint image corresponding to each of the extracted feature points as the first feature amount;
A first feature vector calculation unit that calculates a first feature vector by selecting a plurality of pixels from the first feature image and arranging the first feature obtained from the plurality of selected pixels; ,
A second feature amount image generating unit that generates a second feature amount image by setting a pixel value of the fingerprint image corresponding to each of the extracted feature points as the second feature amount;
A second feature quantity vector calculating unit that calculates a second feature quantity vector by selecting a plurality of pixels from the second feature quantity image and arranging the second feature quantities obtained from the selected plurality of pixels; ,
The fingerprint image processing apparatus according to any one of appendices 1 to 4, further comprising:
[Appendix 6]
The feature vector calculation unit includes:
A first feature amount image smoothing unit that smoothes pixel values of the first feature amount image using pixel values of pixels around the pixel;
A second feature amount image smoothing unit that smoothes pixel values of the second feature amount image using pixel values of pixels around the pixel;
The fingerprint image processing apparatus according to appendix 5, further comprising:
[Appendix 7]
The first feature vector calculating unit selects a pixel at a sampling position defined in a lattice shape from a plurality of pixels forming the first feature image, and acquires the first feature from the pixel at the sampling position And
The second feature quantity vector calculation unit selects a pixel at the sampling position from a plurality of pixels forming the second feature quantity image, and acquires the second feature quantity from the pixel at the sampling position. 6 fingerprint image processing apparatus;
[Appendix 8]
The collation target candidate extraction unit calculates a difference value between elements corresponding to each of the search-side feature quantity vector and the collation-side feature quantity vector, and compares the total value of the calculated difference values with a threshold. The fingerprint image processing apparatus according to any one of appendices 2 to 7, wherein a collation target candidate for a fingerprint image to be searched is extracted.
[Appendix 9]
The feature quantity vector calculation unit further includes a feature quantity vector generation unit that generates the search-side feature quantity vector by combining the first and second feature quantity vectors. Fingerprint image processing device.
[Appendix 10]
The feature vector calculating unit calculates the first and second feature vectors related to the fingerprint image to be collated;
The fingerprint image processing apparatus according to appendix 9, wherein the feature vector generation unit generates the collation-side feature vector by combining the first and second feature vectors related to the fingerprint image to be collated.
[Appendix 11]
The fingerprint image processing apparatus according to any one of appendices 5 to 10, wherein the feature point extraction unit extracts at least one of a ridge end point and a ridge branch point included in the fingerprint image as a feature point.
[Appendix 12]
A plurality of first feature amounts calculated from a rotation direction of the feature point direction with respect to a reference point in the fingerprint image, which is a feature amount of each of the plurality of feature points in the fingerprint image to be searched. Calculating a first feature vector,
Calculating a second feature quantity vector having a plurality of second feature quantities as elements, which is a feature quantity relating to the feature point direction and calculated from the orientation of the feature point direction with respect to the reference point;
A fingerprint image processing method including:
[Appendix 13]
A plurality of first feature amounts calculated from a rotation direction of the feature point direction with respect to a reference point in the fingerprint image, which is a feature amount of each of the plurality of feature points in the fingerprint image to be searched. A process of calculating a first feature vector,
A process for calculating a second feature quantity vector having a plurality of second feature quantities as elements, which is a feature quantity related to the feature point direction and calculated from an orientation of the feature point direction with respect to the reference point;
For causing a computer that controls the fingerprint image processing apparatus to execute the program.
Note that the form of Supplementary Note 12 and Supplementary Note 13 can be expanded to the form of Supplementary Note 2 to the form of Supplementary Note 11, similarly to the form of Supplementary Note 1.

  The disclosure of the cited patent document is incorporated herein by reference. Within the scope of the entire disclosure (including claims) of the present invention, the embodiments and examples can be changed and adjusted based on the basic technical concept. In addition, various combinations or selections of various disclosed elements (including each element in each claim, each element in each embodiment or example, each element in each drawing, etc.) within the scope of the entire disclosure of the present invention. Is possible. That is, the present invention of course includes various variations and modifications that could be made by those skilled in the art according to the entire disclosure including the claims and the technical idea. In particular, with respect to the numerical ranges described in this document, any numerical value or small range included in the range should be construed as being specifically described even if there is no specific description.

DESCRIPTION OF SYMBOLS 10 Fingerprint imaging | photography apparatus 20 Database 30 Collation target extraction apparatus 100 Fingerprint image processing apparatus 101,301,301a, 301b Feature-value vector calculation part 201 Communication part 202 Input part 203 Display part 204 Output part 205 Storage part 206 Collation target determination part 302 Collation Target candidate extraction unit 401 Feature point extraction unit 402 First feature amount image generation unit 403 Second feature amount image generation unit 404 First feature amount image smoothing unit 405 Second feature amount image smoothing unit 406 First feature amount vector calculation unit 407 Second feature vector calculation unit 408 Feature vector generation unit 501-503, 511-513 Ridge 504, 514 Arrow 520-528 Pixel 601-604 Region

Claims (13)

A plurality of first feature amounts calculated from a rotation direction of the feature point direction with respect to a reference point in the fingerprint image, which is a feature amount of each of the plurality of feature points in the fingerprint image to be searched. A first feature vector,
A feature quantity vector calculation unit for calculating a feature quantity related to the feature point direction and a second feature quantity vector having a plurality of second feature quantities as elements, which is calculated from an orientation of the feature point direction with respect to the reference point A fingerprint image processing apparatus.   A search-side feature quantity vector corresponding to the search target fingerprint image generated from the first and second feature quantity vectors, and a feature quantity vector corresponding to the fingerprint image to be collated, the first and second features A collation target for extracting a collation target candidate for the fingerprint image to be searched from among a plurality of fingerprint images to be collated by comparing the elements of the quantity vector with a collation-side feature vector comprising the same type of elements The fingerprint image processing apparatus according to claim 1, further comprising a candidate extraction unit.   The feature quantity vector calculation unit calculates the first feature quantity by quantifying an angle formed by a first reference line connecting the reference point and the feature point and a straight line in the feature point direction. Fingerprint image processing device.   4. The fingerprint image processing according to claim 3, wherein the feature amount vector calculation unit calculates the second feature amount by quantifying an angle formed by a straight line in the feature point direction and a second reference line orthogonal to the first reference line. 5. apparatus. The feature vector calculation unit includes:
A feature point extraction unit for extracting the plurality of feature points from the fingerprint image;
A first feature amount image generating unit that generates a first feature amount image by setting a pixel value of the fingerprint image corresponding to each of the extracted feature points as the first feature amount;
A first feature vector calculation unit that calculates a first feature vector by selecting a plurality of pixels from the first feature image and arranging the first feature obtained from the plurality of selected pixels; ,
A second feature amount image generating unit that generates a second feature amount image by setting a pixel value of the fingerprint image corresponding to each of the extracted feature points as the second feature amount;
A second feature quantity vector calculating unit that calculates a second feature quantity vector by selecting a plurality of pixels from the second feature quantity image and arranging the second feature quantities obtained from the selected plurality of pixels; ,
The fingerprint image processing apparatus according to any one of claims 1 to 4, further comprising: The feature vector calculation unit includes:
A first feature amount image smoothing unit that smoothes pixel values of the first feature amount image using pixel values of pixels around the pixel;
A second feature amount image smoothing unit that smoothes pixel values of the second feature amount image using pixel values of pixels around the pixel;
The fingerprint image processing apparatus according to claim 5, further comprising: The first feature vector calculating unit selects a pixel at a sampling position defined in a lattice shape from a plurality of pixels forming the first feature image, and acquires the first feature from the pixel at the sampling position And
6. The second feature quantity vector calculation unit selects a pixel at the sampling position from a plurality of pixels forming the second feature quantity image, and acquires the second feature quantity from the pixel at the sampling position. Or the fingerprint image processing apparatus of 6. The collation target candidate extraction unit calculates a difference value between elements corresponding to each of the search-side feature quantity vector and the collation-side feature quantity vector, and compares the total value of the calculated difference values with a threshold. The fingerprint image processing apparatus according to claim 2, wherein a collation target candidate for a fingerprint image to be searched is extracted. The fingerprint image processing apparatus according to claim 2, wherein the feature vector calculation unit further includes a feature vector generation unit that generates the search-side feature vector by combining the first and second feature vectors. . The feature vector calculating unit calculates the first and second feature vectors related to the fingerprint image to be collated;
The fingerprint image processing apparatus according to claim 9, wherein the feature vector generation unit generates the collation-side feature vector by combining the first and second feature vectors related to the fingerprint image to be collated. The feature point extraction unit extracts the at least one branch point of the end points and ridge ridge included in the fingerprint image as a feature point, the fingerprint image processing apparatus according to any one of claims 5 to 7 . A plurality of first feature amounts calculated from a rotation direction of the feature point direction with respect to a reference point in the fingerprint image, which is a feature amount of each of the plurality of feature points in the fingerprint image to be searched. Calculating a first feature vector,
Calculating a second feature quantity vector having a plurality of second feature quantities as elements, which is a feature quantity relating to the feature point direction and calculated from the orientation of the feature point direction with respect to the reference point;
A fingerprint image processing method including: A plurality of first feature amounts calculated from a rotation direction of the feature point direction with respect to a reference point in the fingerprint image, which is a feature amount of each of the plurality of feature points in the fingerprint image to be searched. A process of calculating a first feature vector,
A process for calculating a second feature quantity vector having a plurality of second feature quantities as elements, which is a feature quantity related to the feature point direction and calculated from an orientation of the feature point direction with respect to the reference point;
For causing a computer that controls the fingerprint image processing apparatus to execute the program.