JPWO2017038668A1 - Image processing apparatus, image processing method, and program - Google Patents

Abstract

The present invention provides an image processing apparatus that calculates a feature amount that contributes to accuracy improvement in collation of fingerprints and the like. The image processing apparatus includes an image input unit, a core line image extraction unit, and a core line addition unit. The image input unit inputs an image in which a curved stripe pattern is formed by ridges. The core line image extraction unit generates a core line image obtained by extracting a core line from the image. The core line adding unit detects a ridge or area having a predetermined characteristic from the core line image, and adds the core line to the detected ridge or area.

Description

[Description of related applications]
The present invention is based on a Japanese patent application: Japanese Patent Application No. 2015-168471 (filed on August 28, 2015), and the entire contents of this application are incorporated in the present specification by reference.
The present invention relates to an image processing apparatus, an image processing method, and a program. In particular, the present invention relates to an image processing apparatus, an image processing method, and a program for processing image data having a curved stripe pattern such as a fingerprint image.

  Fingerprints and palm prints composed of a large number of ridges with a curved stripe pattern have long been used as a means for confirming a person. In particular, verification using a residual fingerprint left at the crime scene is an effective investigation means. Many police agencies and the like have introduced a fingerprint collation system (AFIS) using a computer (Automatic Fingerprint Identification System). In the fingerprint verification system, the fingerprint image registered in the database is compared with the feature points (also called Minutia) of each retained fingerprint collected at the crime scene, and the person corresponding to the retained fingerprint is identified. The

  In many cases, an end point or a branch point of a fingerprint ridge is used as a feature point (fingerprint feature point) used for fingerprint collation. For example, as disclosed in “4.3 Minute-Based Methods” of Non-Patent Document 1, feature point matching using the end points and branch points of fingerprint ridges is used.

  Also, in the operation of fingerprint verification systems in police agencies, for high-quality imprinted fingerprints, feature points are automatically extracted and registered in the database, but for low-quality fingerprint images In many cases, Examiner manually inputs feature points. In addition, the feature points of the fingerprint image registered in the database are treated as data on the search side (file side), and the feature points input by the inspector are handled as data on the search side (search side). Fingerprint collation (person identification) is performed based on the collation score.

  Non-Patent Document 2 describes such descriptive terms as core, delta, and innermost recurve or loop.

  In Patent Document 1, a feature point direction is described as a feature amount that characterizes the direction of a feature point (branch point, end point).

JP 2010-267016 A

D.Maltoni, “Handbook of Fingerprint Recognition”, Springer, 2003 United States. Federal Bureau of Investigation, “The Science of Fingerprints-Classification and Uses”, 1990

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

  In many cases, not only the position of the end point and the branch point but also the feature amount related to the direction as disclosed in Patent Document 1 is used as the feature amount characterizing the fingerprint image. For example, when feature points (end points, branch points) are extracted from the fingerprint image shown in FIG. 45 (a), the result is as shown in FIG. 45 (b). In the drawings to be referred to in the following description, the end points of the ridges are indicated by circles and the branch points are indicated by squares.

  As shown in FIG. 45 (b), three end points and branch points are extracted from the fingerprint image of FIG. 45 (a). In addition, for each end point and branch point, a feature amount regarding the direction of each feature point is calculated. In FIG. 45 (b), a short line extending from each of the end points and the branch points indicates the calculated feature amount regarding the direction of the feature point.

  The position and direction of each feature point (value obtained by digitizing the direction of the short line shown in FIG. 45B) is often calculated as a feature amount (feature amount vector) that characterizes each fingerprint image. Since many end points and branch points are scattered throughout the fingerprint image, the position and direction of each feature point scattered in the fingerprint image is calculated as a feature amount. However, as a result of investigations by the inventors, it has been found that a feature amount related to the direction of a feature point (hereinafter referred to as a feature point direction) cannot be stably calculated in a part of the fingerprint image. Specifically, it has been found that in the regions such as the core region and the delta region of the fingerprint image, the feature point direction cannot be calculated stably from the feature points included in these regions.

  FIG. 46 is a diagram illustrating a partial region of a fingerprint image. In FIG. 46, a core region 701 and a delta region 702 are shown. An area 703 is an area other than the core area and the delta area. 46, it can be seen that the fluctuation in the ridge direction in the core region 701 and the delta region 702 is much larger than that in the other regions 703. Specifically, in the region 703, it can be confirmed that all the ridges flow in substantially the same direction from the upper left to the lower right, whereas in the core region 701, the direction of the ridge greatly changes to an acute angle. . Moreover, in the delta area | region 702, it can confirm that each ridge branches in three directions starting from the center point.

  Although details will be described later, when calculating the feature point direction of the branch point, the angle difference at the three inner angles formed by the three ridges is used. In the regions other than the core region and the delta region, the angle difference is large, and even if the ridge direction fluctuates due to the influence of noise or the like, the influence on the angle difference is small. As a result, the feature point direction is stably calculated from the feature points existing in the regions other than the core region and the delta region.

  However, as shown in FIG. 46, the core region and the delta region are regions in which the ridge direction can be greatly changed. The relationship between the lines can easily change. As a result, the angle difference at the three inner angles formed by the three ridges also varies, and the feature point direction calculated based on the angle difference is not stable. As described above, there is a problem that even if a high-quality imprinted fingerprint is used, the feature point direction cannot be stably calculated in the core region and the delta region.

  If the feature point direction, which is one of the feature quantities that characterize a fingerprint image, cannot be calculated stably, even if it is a fingerprint image (versus-fingerprint image) of the same person taken at different times, places, etc., the corresponding feature The feature point directions of the points will be different. As a result, in the matching of two fingerprint images, the feature points that are originally paired are determined to be non-paired feature points, and the matching score deteriorates (matching accuracy deteriorates).

  From such a situation, when the feature point direction is calculated from the fingerprint image related to the retained fingerprint, there is a case where the feature point included in the core region or the delta region is not used. However, since various ridge shapes exist in the core region and the delta region and have shapes specific to individual fingerprints, it can be said that these regions are originally effective regions for fingerprint matching. In other words, the core region and delta region are not fully utilized because the feature point direction cannot be calculated stably, and feature quantities that appropriately characterize the feature points included in these regions are calculated. By doing so, it can be said that there is room for improving the accuracy of fingerprint matching.

  An object of the present invention is to provide an image processing apparatus, an image processing method, and a program for calculating a feature amount that contributes to accuracy improvement in fingerprint collation.

  According to a first aspect of the present invention, an image input unit that inputs an image in which a curved stripe pattern is formed by ridges, a core line image extraction unit that generates a core line image obtained by extracting a core line from the image, There is provided an image processing apparatus comprising: a core line adding unit that detects a ridge or area having a predetermined characteristic from the core line image and adds a core line to the detected ridge or area.

  According to a second aspect of the present invention, from the step of inputting an image in which a curved stripe pattern is formed by ridges, the step of generating a core line image obtained by extracting a core line from the image, and the core line image, Detecting a ridge or region having a defined characteristic, and adding a core line to the detected ridge or region.

According to a third aspect of the present invention, from a process of inputting an image in which a curved stripe pattern is formed by ridges, a process of generating a core line image obtained by extracting a core line from the image, and the core line image, There is provided a program for detecting a ridge or area having a defined characteristic and adding a core line to the detected ridge or area, and causing a computer mounted on the image processing apparatus to execute the process.
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, there are provided an image processing apparatus, an image processing method, and a program that contribute to calculating a feature amount that contributes to accuracy improvement in collation of fingerprints and the like.

It is a figure for demonstrating the outline | summary of one Embodiment. It is a figure which shows an example of a structure of the fingerprint collation system which concerns on 1st Embodiment. It is a figure which shows an example of the internal structure of the feature-value production | generation apparatus which concerns on 1st Embodiment. It is a figure which shows an example of a core line image. It is a figure for demonstrating calculation of the feature point direction regarding a branch point. It is a figure for demonstrating the calculation of the feature point direction regarding an end point. It is a figure for demonstrating the numerical value of a feature point direction. It is a figure for demonstrating the problem which concerns on the calculation of the feature point direction in a core area | region. It is a figure which shows an example of a remains fingerprint image. It is a figure for demonstrating the shift | offset | difference of a feature point direction. 5 is a flowchart illustrating an example of an operation of a feature amount calculation unit 13. It is a figure which shows an example of the 1st and 2nd feature point direction. It is a figure which shows an example of the 1st and 2nd feature point direction. It is a figure for demonstrating the calculation of the 2nd feature point direction from the branch point which exists in a delta area | region. It is a figure for demonstrating the calculation process of the 2nd feature point direction regarding an end point. It is a figure which shows an example of the 1st and 2nd feature point direction of an end point. It is a figure for demonstrating the calculation of the 2nd feature point direction by the feature-value calculation part. It is a figure for demonstrating the calculation of the 2nd feature point direction by the feature-value calculation part. It is a figure which shows an example of the feature-value which a feature-value calculation part calculates. It is a figure which shows an example of the internal structure of the input device which concerns on 1st Embodiment. It is a figure which shows an example of the operation screen displayed by an input device. It is a figure which shows an example of the internal structure of the collation apparatus which concerns on 1st Embodiment. It is a figure for demonstrating the collation process by a feature point collation part. It is a figure for demonstrating the operation | movement which detects the correspondence between the feature points by a feature point collation part. It is a figure for demonstrating the operation | movement which detects the correspondence between the feature points by a feature point collation part. It is a figure for demonstrating the operation | movement which detects the correspondence between the feature points by a feature point collation part. It is a figure for demonstrating the operation | movement which detects the correspondence between the feature points by a feature point collation part. It is a figure for demonstrating the correspondence between two feature points. It is a figure for demonstrating the correspondence between two feature points. It is a figure which shows an example of the collation result which a collation apparatus outputs. It is a sequence diagram which shows an example of the whole operation | movement of the fingerprint collation system which concerns on 1st Embodiment. It is a figure for demonstrating a core hoof line. It is a figure for demonstrating that a feature point may not exist in a delta area | region. It is a figure which shows an example of an internal structure of the feature-value production | generation apparatus which concerns on 2nd Embodiment. It is a figure for demonstrating the core wire addition part which concerns on 2nd Embodiment. It is a figure for demonstrating the core wire addition part which concerns on 2nd Embodiment. It is a figure for demonstrating the input of the protrusion core wire by the inspector. It is a figure for demonstrating the input of the delta dot core wire by an inspector. It is a figure for demonstrating transition of a core hoof line. It is a figure which shows an example of the fingerprint image in a delta area | region. It is a figure which shows an example of an internal structure of the image processing apparatus which concerns on one Embodiment. It is a figure which shows an example of the hardware constitutions of a feature-value production | generation apparatus. It is a figure which shows an example of the hardware constitutions of an input device. It is a figure which shows an example of the hardware constitutions of a collation apparatus. It is a figure for demonstrating the feature point and feature-value of a fingerprint image. It is a figure which shows the one part area | region of a fingerprint image.

  First, an outline of one embodiment will be described. 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.

  An image processing apparatus 100 according to an embodiment includes an image input unit 101, a core line image extraction unit 102, and a core line addition unit 103. The image input unit 101 inputs an image in which a curved stripe pattern is formed by ridges. The core line image extraction unit 102 generates a core line image obtained by extracting a core line from the image. The core line adding unit 103 detects a ridge or area having a predetermined characteristic from the core line image, and adds the core line to the detected ridge or area.

  Although details will be described later, there are rare cases where feature points do not exist in the core region and the delta region. In addition, the core region and the delta region are characterized by being easily affected by noise and the like. Due to such a situation, in the fingerprint images on the search side and the file side, a feature point may exist in one core region or the like, and a feature point may not exist in the other core region or the like. When such a situation occurs, high collation accuracy cannot be expected. Therefore, the image processing apparatus 100 according to an embodiment detects ridges (core hoof lines described later) and delta regions that appear in the core region from the core image, and adds the core wire to these detection targets. As a result, in two fingerprint images (core line images), it is possible to avoid a situation in which a feature point exists on one side and a feature point does not exist on the other side, and deterioration of matching accuracy can be suppressed.

  Hereinafter, specific embodiments will be described in more detail with reference to the drawings. In addition, in each embodiment, the same code | symbol is attached | subjected to the same component and the description is abbreviate | omitted.

[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 configuration of the fingerprint matching system according to the first embodiment. Referring to FIG. 2, the fingerprint collation system includes a feature quantity generation device 10, a database 20, an input device 30, and a collation device 40.

  The feature quantity generation device 10 inputs data relating to a fingerprint image and its ID (Identifier) information, and generates a feature quantity (feature quantity vector) that characterizes the inputted fingerprint image. The feature amount generation device 10 outputs the generated feature amount together with the fingerprint image ID information to the database 20.

  The database 20 stores the fingerprint image feature quantity and its ID information in association with each other.

  The input device 30 is, for example, a device used by an expert. The input device 30 is a device that accepts an operation by a pointing device such as a mouse or a tablet, for example, and an expert officer inputs a feature amount. The input device 30 outputs the feature amount input by the inspector to the verification device 40 together with the ID information of the fingerprint image.

  The collation device 40 executes a collation process using the feature amount input by the expert (a feature amount output from the input device 30). Specifically, the collation device 40 compares the feature amount input by the inspector with the feature amount obtained by accessing the database 20, thereby the feature that is the same as or similar to the feature amount input by the inspector. Identify fingerprint images with volume.

  Here, the imprint fingerprint is a fingerprint collected for the purpose of registration in a database or the like, and has a feature that the area of the ridge region is wide and the quality is good. On the other hand, a deceased fingerprint is a fingerprint detained at a crime scene or the like, and has a feature that the area of a clear ridge region is narrow and the quality is often low. In order to obtain a highly reliable feature quantity, a forensic officer often inputs a feature quantity for a remnant fingerprint.

  In view of such circumstances, in the first embodiment, the feature amount generation device 10 inputs an image related to a fingerprint, and the fingerprint image that the inspector inputs the feature amount in the input device 30 is an image related to a retained fingerprint. And However, this is not intended to limit the fingerprint images that can be handled by each device. The feature quantity generation device 10 may generate a feature quantity from the retained fingerprint and provide the feature quantity to the collation device 40, or an inspector may extract the feature quantity from the imprinted fingerprint. In other words, the fingerprints to be collated may be either imprinted fingerprints or retained fingerprints. However, in the criminal investigation use, the above-described configuration is adopted in the first embodiment in consideration of the fact that the remnant fingerprint and the fingerprint are often collated.

[Configuration and Operation of Feature Generation Device]
FIG. 3 is a diagram illustrating an example of an internal configuration of the feature value generation apparatus 10. Referring to FIG. 3, the feature quantity generation device 10 includes a fingerprint image input unit 11, a core line image extraction unit 12, a feature quantity calculation unit 13, a feature quantity output unit 14, and a storage unit 15. Is done. Each unit such as the fingerprint image input unit 11 is configured to be able to exchange data with each other, and can access data stored in the storage unit 15.

  The fingerprint image input unit 11 is a means for inputting data relating to a fingerprint image (an image in which a curved stripe pattern is formed by ridges) and ID information of the fingerprint image from the outside. For example, the fingerprint image input unit 11 takes in digital data (image file) of a fingerprint image stored in an external storage medium such as a USB (Universal Serial Bus) memory, and delivers the data to the core image extraction unit 12. Alternatively, the fingerprint image input unit 11 may input data relating to a fingerprint image via a network. Alternatively, the digitized fingerprint image may be acquired by mounting the scanner function in the fingerprint image input unit 11 instead of inputting the digitized fingerprint image by a scanner or the like.

  There is a standardized standard for fingerprint images. Specifically, there are ANSI / NIST-ITL-1-2000 Data Format for the Interchange of Fingerprint, Facial, & Scar Mark & Tattoo (SMT) Information standardized by the National Institute of Standards and Technology. It is desirable that the fingerprint image input unit 11 can handle a digitized fingerprint image (for example, a fingerprint image having a resolution of 500 dpi) based on the above standard.

  The storage unit 15 is configured by a medium such as a RAM (Random Access Memory), a ROM (Read Only Memory), or an HDD (Hard Disk Drive), for example, and each unit such as the core image extraction unit 12 stores a work area and data. Use as

  The core line image extraction unit 12 is means for extracting a core line image from the acquired fingerprint image. For example, the core image extraction unit 12 extracts a core image as shown in FIG. Note that the core image extraction unit 12 can use the core image extraction method disclosed in “3 Fingerprint Analysis and Representation” of Non-Patent Document 1. Therefore, although detailed description regarding the extraction of the core line image is omitted, the core line image extraction unit 12 extracts the core line image in the following general procedure.

  The core line image extraction unit 12 first extracts the direction of the ridge of the fingerprint image. Thereafter, the core image extraction unit 12 emphasizes each ridge along the ridge direction and generates a binary image. Thereafter, the core image extraction unit 12 extracts a core image (core data) by converting the binary image into a core.

  The feature amount calculation unit 13 is a means for extracting feature points from the core image and calculating feature amounts that characterize the feature points. The feature amount calculated by the feature amount calculation unit 13 includes the position of the feature point (end point, branch point) and the feature point direction that characterizes the feature point according to the direction.

  The feature quantity calculation unit 13 extracts feature points by extracting branch points and end points of the core line from the core line image extracted by the core line image extraction unit 12. Note that the feature point extraction method disclosed in “3 Fingerprint Analysis and Representation” of Non-Patent Document 1 can be used as a procedure for extracting feature points from a core image. Therefore, the detailed description regarding feature point extraction is abbreviate | omitted. In addition, the feature amount calculation unit 13 extracts a singular point (core type singular point or delta type singular point) of a fingerprint image and extracts it as one of the feature amounts.

  For example, the feature quantity calculation unit 13 extracts branch points 201 and 202 and end points 211 and 212 as shown in FIG. 4 as feature points, and stores the position of each feature point and its type (branch point and end point). 15 to register.

  The feature point direction calculation process that is normally performed prior to the feature point direction calculation by the feature amount calculation unit 13 and its problems will be described below.

  First, calculation of the feature point direction regarding the branch point will be described. At this time, calculation of the feature point direction 221 of the branch point 203 shown in FIG. First, the three end points 231 to 233 are determined by tracing each of the three ridges (core lines) forming the branch point 203 by a predetermined distance from the branch point 203 (tracing back on the ridge line) (FIG. 5). (See (b)).

  Next, three angles a1 to a3 formed by three straight lines determined by the branch point 203 and the three end points 231 to 233 are calculated (see FIG. 5C). Next, of the three calculated angles, the smallest angle is determined, and the direction that bisects the angle is calculated as the feature point direction 221 (see FIG. 5B).

  In the present disclosure, of the two angles formed by two straight lines starting from one point, the smaller angle is defined as the inner angle, and the larger angle is defined as the outer angle. Of the three angles formed by three straight lines starting from one point, the one with the smallest angle is the smallest interior angle, the one with the second smallest angle is the second smallest interior angle, The larger one is defined as the maximum interior angle. In the example of FIG. 5C, if the angle a1 is the minimum interior angle and the angle a2 is smaller than the angle a3, the angle a2 is the second minimum interior angle.

  Next, calculation of the feature point direction regarding the end points will be described with reference to FIG. First, the end point 234 is calculated by tracing a certain distance from the end point 213 on the ridge (core line) forming the end point. The direction from the end point 213 toward the end point 234 is calculated as the feature point direction of the end point 213. Note that a feature point direction including first and second feature point directions, which will be described later, is formed by an X axis in a two-dimensional coordinate system in a fingerprint image and a straight line by the feature point direction, as shown in FIG. It is expressed as a feature value using the angle θ.

  The trace distance when calculating the branch point and the end point is an important parameter that greatly affects the calculated feature point direction. In the first embodiment, two ridge intervals are set as the trace distance. However, the trace distance is not limited to this.

  The feature point direction of the feature point is calculated by the method described with reference to FIG. 5 and FIG. 6, but the calculated feature point is calculated in the core region (near the fingerprint center point) and the delta region (delta region). There is a problem that the direction becomes unstable. The reason will be described with reference to FIG. Note that the fingerprint image shown in FIG. 8A is an excerpt of the core region.

  When the ridges forming the branch point 204 shown in FIG. 8A are traced in a predetermined distance in three directions according to the above method, three terminal points 235 to 237 are obtained (see FIG. 8B). ). Next, among the three inner angles a1 to a3 formed by three straight lines connecting the branch point 204 and the three end points 235 to 237, the direction that bisects the minimum inner angle a1 is a feature point of the branch point 204. It is calculated as a direction (see FIG. 8C).

  Here, referring to FIG. 8B, it is understood that the difference between the minimum internal angle a1 and the second minimum internal angle a2 is small among the three internal angles. This means that the relationship between the minimum interior angle and the second minimum interior angle can be reversed if the various factors that define the three interior angles vary even slightly. For example, even in the same fingerprint image, there is a possibility that the direction of the feature point may vary greatly due to a slight difference in binarization and core processing applied to the fingerprint image.

  FIG. 9 is a diagram illustrating an example of a remnant fingerprint image. FIGS. 10A and 10D are enlarged views of the periphery of the branch point 205 of the fingerprint image shown in FIG. 10A and 10D, the fingerprint image is subjected to different types of core processing, and the results are displayed in an overlapping manner. FIG. 10B is a diagram in which the core wire is extracted from FIG. 10A, and FIG. 10E is a diagram in which the core wire is extracted from FIG. 10D. When the above-described processing related to the calculation of the feature point direction (processing for calculating the direction that bisects the minimum internal angle) is applied to the core image shown in each of FIGS. 10B and 10E, FIG. The characteristic point directions 242 and 243 shown in FIG. 10C and FIG. 10F are obtained.

  Referring to FIG. 10, it can be confirmed that the direction of the feature point of the branch point existing in the core region varies greatly due to a slight difference in the core line even with the same residual fingerprint. In this way, it is difficult to stably and uniquely determine the feature point direction in the core region, and even for fingerprint images taken from the same person, different feature point directions are calculated on the search side and the file side. May end up. As a result, the difference in the feature point direction between the paired feature points on the search side and the file side becomes large, which can be a factor that lowers the matching score and causes pairing failure.

  Although the above description is for a case where the core line processing method is different, the phenomenon that the feature point direction is not uniquely determined in the core region can easily occur due to noticeable image distortion, dirt, or the like in the retained fingerprint.

  Furthermore, in the delta region, it can be said that the angle difference between the three interior angles formed by the straight lines connecting the branch point and the three end points is smaller than the branch point of the core region. In other words, since the three ridges extend in three directions in the delta region, the three interior angles are approximately 120 degrees. Therefore, in the delta region, the calculated feature point direction tends to be more unstable.

  Further, as described above, the quality of a retained fingerprint is often low and is not suitable for automatic extraction of feature points by information processing using a computer. For this reason, a remnant fingerprint specialist (Lent Examiner) often manually inputs the position of the feature point and the direction of the feature point. Such an input operation is called “feature point manual input”. Even a remnant fingerprint specialist with specialized knowledge cannot easily determine the minimum interior angle at the bifurcation point of the core region or delta region. Different remnant fingerprint specialists specify different feature point directions. Often done. As a result, the feature values on the search side and the file side are different, which is one of the causes of deterioration of collation accuracy.

  In view of the above situation, in the feature quantity generation device 10 according to the first embodiment, the feature point direction according to the feature point direction calculation method described with reference to FIGS. 5 and 6 is different from the method. The feature point direction by the method is calculated.

  In the following description, the feature point direction calculation method described with reference to FIGS. 5 and 6 is referred to as a first calculation method, and the feature point direction calculated by the method is referred to as a first feature point direction. write. Further, a method different from the first calculation method described below is referred to as a second calculation method, and a feature point direction calculated by the method is referred to as a second feature point direction (Secondary Direction).

  FIG. 11 is a flowchart illustrating an example of the operation of the feature amount calculation unit 13. With reference to FIG. 11, a second calculation method mainly related to the feature point direction will be described.

  In step S <b> 101, the feature amount calculation unit 13 extracts feature points from the acquired core image, and registers the position and type in the storage unit 15. Thereafter, the feature amount calculation unit 13 calculates the feature point direction of each feature point scattered in the core line image by executing the processing after step S102.

  In step S <b> 102, the feature amount calculation unit 13 confirms whether the feature point that is the target of feature point direction calculation is a branch point or an end point.

  In the case of a branch point (step S102, Yes branch), the feature amount calculation unit 13 executes the processes according to steps S103 to S105. In the case of the end point (step S102, No branch), the feature amount calculation unit 13 executes the processing according to steps S106 to S116.

  In step S103, the feature amount calculation unit 13 calculates the first feature point direction of the branch point by the first calculation method. Here, it is assumed that the core image shown in FIG. 8B and the three inner angles a1 to a3 are obtained by the first calculation method, and the first feature point direction shown in FIG. 8C is calculated.

  When the calculation of the first feature point direction ends, the feature amount calculation unit 13 calculates a difference value (angle difference) between the second minimum interior angle and the minimum interior angle, and determines whether the difference value is equal to or less than a predetermined threshold value. (Step S104). In the example shown in FIG. 8B, the difference value is calculated by subtracting the angle of the minimum internal angle a1 from the second minimum internal angle a2.

  If the difference value is equal to or less than the threshold value (step S104, Yes branch), the feature amount calculation unit 13 calculates the second feature point direction of the branch point (step S105). Specifically, the feature amount calculation unit 13 calculates the direction that bisects the second minimum interior angle as the second feature point direction. For example, as shown in FIG. 12, in addition to the first feature point direction 301, a second feature point direction 311 that bisects the second minimum interior angle a2 is calculated. In addition, the threshold value (threshold value in step S104) used for determining whether or not the feature amount calculation unit 13 calculates the second feature point direction can be, for example, about 20 degrees to 30 degrees.

  In this disclosure, when the first feature point direction is illustrated, a solid line with a bulge is used, and when the second feature point direction is illustrated, a solid line without a bulge is used.

  The feature amount calculation unit 13 calculates the direction that bisects the second minimum internal angle as the second feature point direction when the angle difference between the second minimum internal angle and the minimum internal angle is small (for example, the angle difference is 30 degrees or less). . When the difference value is larger than the threshold value (step S104, No branch), the feature amount calculation unit 13 ends the process without calculating the second feature point direction for the branch point.

  FIG. 13A is a diagram illustrating an example of the first feature point direction and the second feature point direction calculated from the fingerprint image illustrated in FIG. 10A, and FIG. 13B is a diagram illustrating FIG. It is a figure which shows an example of the 1st feature point direction calculated from the fingerprint image shown, and a 2nd feature point direction. Referring to FIG. 13, the first feature point direction 302 shown in FIG. 13A and the second feature point direction 313 shown in FIG. 13B match (substantially match; the deviation between the two directions is within a predetermined range) It can be seen that the second feature point direction 312 in FIG. 13A and the first feature point direction 303 in FIG. 13B match (substantially match).

  Although collation using these feature point directions will be described later, by introducing the second feature point direction as a new feature amount, it is possible to suppress degradation of collation accuracy caused by variation in the feature point direction in the core region or the like. . That is, even if the first feature point direction is different between two feature points, if one first feature point direction and the other second feature point direction coincide, the two feature points are paired feature points (pairs). Characteristic point).

  The second calculation method described above is a case where the branch point existing in the core region is taken as an example. However, for the branch point existing in the delta region, the direction of the second feature point from the branch point existing in the core region is also determined. It can be calculated in the same way as the calculation.

  For example, referring to FIG. 14, the first feature point direction 304 is calculated by the minimum inner angle a1, and the second feature point direction 315 is calculated by the second minimum inner angle a2 as the feature point directions characterizing the branch point 321 existing in the delta region. Is done.

  As described above, when the extracted feature point is a branch point, the feature amount calculation unit 13 selects one of the three interior angles formed by the three ridges forming the branch point (for example, Based on the minimum inner angle a1) shown in FIG. 8B, the first feature point direction is calculated. Further, the feature amount calculation unit 13 is based on a relationship between at least two of the three inner angles (for example, a magnitude relationship between the minimum inner angle a1 and the second minimum inner angle a2 illustrated in FIG. 8B). The second feature point direction is calculated. Specifically, the feature amount calculation unit 13 determines the second feature point when the angle difference between the inner angle having the smallest angle and the inner angle having the second smallest angle is smaller than the threshold among the three inner angles. Calculate the direction. More specifically, the feature amount calculation unit 13 divides the minimum interior angle into two, the direction toward the terminal point obtained by tracing the ridge is the first feature point direction, and the second minimum interior angle is 2 A direction toward the terminal point obtained by tracing the ridge is defined as a second feature point direction.

  When the feature point is an end point (step S102 in FIG. 12, No branch), the feature amount calculation unit 13 calculates the first feature point direction of the end point by the first calculation method (step S106).

  Next, the feature quantity calculation unit 13 calculates a candidate for the second feature point direction of the end point (step S107). Specifically, the feature amount calculation unit 13 determines a distance (hereinafter referred to as a second trace distance) longer than a normal trace distance from the end point (a trace distance during the first calculation method; hereinafter referred to as a first trace distance). ) Trace on the ridge and calculate two end points by two trace distances. Note that the second trace distance is preferably about twice the first trace distance. For example, referring to FIG. 15, the end point 341 is calculated by the first calculation process, and the end point 342 located at the second trace distance from the end point 331 is calculated. The feature amount calculation unit 13 sets a direction from the end point 331 toward the terminal point 342 according to the second trace distance as a candidate for the second feature point direction.

  In step S108 in FIG. 11, the feature amount calculation unit 13 calculates an interior angle b1 formed from a straight line connecting the end point 331 and the end point 341 and a straight line connecting the end point 331 and the end point 342.

  Thereafter, the feature amount calculation unit 13 compares the calculated angle (inner angle b1 shown in FIG. 15) with a threshold (for example, 45 degrees) (step S109).

  If the calculated interior angle is larger than the threshold (step S109, Yes branch), the feature amount calculation unit 13 determines the second feature point direction candidate set in step S107 as it is as the second feature point direction ( Step S110). In other words, the feature amount calculation unit 13 changes the end point obtained by the second trace distance from the end point when the interior angle obtained from the two end points calculated corresponding to the two trace distances is larger than the threshold value. Let the direction of the straight line which goes to be a 2nd feature point direction.

  Referring to FIG. 16, for example, it is assumed that the end point 343 is obtained by tracing according to the first trace distance, and the end point 344 is obtained by tracing according to the second trace distance. In this case, the direction from the end point 332 to the end point 343 is the first feature point direction, and the direction from the end point 332 to the end point 344 is the second feature point direction.

  As described above, when the extracted feature point is an end point, the feature amount calculation unit 13 goes from the end point to the first end point obtained by tracing the first distance on the ridge that forms the end point. Let the direction be the first feature point direction. Further, the feature amount calculation unit 13 calculates a second terminal point obtained by tracing a second distance longer than the first distance on the ridge that forms the terminal point, and connects the terminal point and the first terminal point. When the internal angle formed at the end point by the straight line and the straight line connecting the end point and the second end point satisfies a predetermined condition (for example, the internal angle by the two straight lines is equal to or greater than the threshold value), The direction toward the end point is calculated as the second feature point direction.

  If the second feature point direction cannot be determined by the method of step S110 (step S109, No branch), the feature amount calculation unit 13 tries to calculate the second feature point direction by the method after step S111.

  FIG. 17 is a diagram for explaining calculation of the second feature point direction by the feature amount calculation unit 13. With reference to FIG. 17, the operation of the feature quantity calculation unit 13 after step S111 will be described.

  In step S111, the feature amount calculator 13 has a predetermined length (for example, a ridge) in a direction (opposite direction; direction reversed 180 degrees) opposite to the first feature point direction 351 based on the first trace distance from the end point 333. An imaginary line 361 having an interval is set.

  Next, the feature quantity calculation unit 13 determines whether or not an intersection of the virtual line 361 and the ridge exists (step S112). In the example of FIG. 17, an intersection point 371 is detected.

  When the intersection of the virtual line and the ridge is detected (step S112, Yes branch), the feature amount calculation unit 13 makes a certain distance in two directions on the ridge from the detected intersection (for example, the ridge interval 2). This is traced, and two end points 345 and 346 are calculated (step S113).

  Next, the feature amount calculation unit 13 includes a straight line connecting the intersection point 371 and the end point 333 (part of the virtual line 361), a straight line connecting the intersection point 371 and the terminal point 345, and a straight line connecting the intersection point 371 and the terminal point 346. The three inner angles a4 to a6 formed at the intersection 371 by the three straight lines are calculated (step S114; see FIG. 17B).

  Thereafter, the feature quantity calculation unit 13 excludes the inner angle a6 having the largest angle among the three inner angles, and the larger inner angle (the inner angle a5 in the example of FIG. 17; the second smallest inner angle) of the two inner angles a4 and a5. ) Is less than or equal to a threshold value (for example, 145 degrees) (step S115).

  When the second minimum interior angle is equal to or smaller than the threshold (step S115, Yes branch), the feature amount calculation unit 13 determines the direction from the intersection point 371 to the end point 345 as the second feature point direction (step S116).

  Referring to FIG. 17C, a first feature point direction 351 and a second feature point direction 352 are calculated as the feature point directions corresponding to the end point 333. In FIG. 17C, the intersection point 371 is moved to the position of the end point 333, and the start point of the second feature point direction 352 is shown to overlap the end point 333.

  When the second minimum interior angle is larger than the threshold value (step S115, No branch), the feature amount calculation unit 13 determines that there is no second feature point direction corresponding to the end point (ends the process).

  When the above calculation is executed, in the region other than the core region and the delta region, the second minimum interior angle (in the example of FIG. 17B, the interior angle a5) is 180 degrees among the angles formed by the three straight lines. It becomes a close value (exceeds 145 degrees defined as a threshold). Therefore, in regions other than the core region and the delta region, there are few end points at which the second feature point direction is calculated.

  In the calculation of the second feature point direction of the end point described with reference to the flowchart of FIG. 11, the calculation of the second feature point direction in step S110 and the second feature point direction in step S116 are performed according to predetermined conditions. Of course, it is possible to execute the calculation of the second feature point direction with respect to the end point.

  The calculation method of the second feature point direction of the end point described with reference to FIGS. 16 and 17 is the description in the core region, but the second feature point direction can also be calculated by the same method for the end point of the delta region. .

  For example, referring to FIG. 18, an end point 334 is present in the delta region. The feature amount calculation unit 13 calculates the end point 347 from the end point 334 by the first trace distance (calculates the first feature point direction). In this case, the interior angle composed of the straight line connecting the end point and the end point by the first trace distance and the straight line connecting the end point and the end point by the second trace distance is smaller than the threshold (step S109, No branch), and step S111. The subsequent processing is executed.

  The feature amount calculation unit 13 sets a virtual line 362 in a direction opposite to the first feature point direction, and obtains an intersection 372 between the virtual line 362 and the ridge (step S111). Next, the feature amount calculation unit 13 traces a certain distance in two directions on the ridge from the intersection point 372, and calculates two end points 348 and 349 (step S113). Next, the feature amount calculation unit 13 is formed by three straight lines including a straight line connecting the intersection point 372 and the end point 334, a straight line connecting the intersection point 372 and the terminal point 348, and a straight line connecting the intersection point 372 and the terminal point 349. The three inner angles a4 to a6 to be calculated are calculated (step S114; see FIG. 18B). In this case, since the second minimum interior angle a5 can be said to be equal to or less than the threshold value, the direction from the intersection point 372 to the terminal point 348 is calculated as the second feature point direction 353 (step S116; see FIG. 18C). 18C, the intersection point 372 is moved to the position of the end point 334, and the start point of the second feature point direction 353 is shown to overlap the end point 334.

  As described above, when the feature amount calculation unit 13 cannot calculate the second feature point direction in step S110, the feature amount calculation unit 13 extends the virtual line starting from the end point in the direction opposite to the first feature point direction. Set. And the feature-value calculation part 13 calculates a 2nd feature point direction, when a virtual line has a ridge and an intersection within a predetermined range, and satisfy | fills a predetermined condition.

  Thus, the feature point feature amount calculation processing by the feature amount calculation unit 13 ends.

  When the feature amount calculation processing by the feature amount calculation unit 13 ends, a feature amount as shown in FIG. 19 is obtained. Referring to FIG. 19, the feature amount calculation unit 13 calculates a feature amount (a feature point type, a position, a first feature point direction, and a second feature point direction) related to feature points existing in the fingerprint image for each fingerprint image. To do. As described above, the second feature point direction is calculated in addition to the first feature point direction in the core region and the delta region, but the second feature point direction is often not calculated in other regions. In FIG. 19, the direction of the second feature point that has not been calculated is described using the symbol “−”.

  The feature amount calculation unit 13 delivers the calculated feature amount (result as shown in FIG. 19) to the feature amount output unit 14.

  The feature amount output unit 14 outputs the feature amount calculated by the feature amount calculation unit 13 to the database 20 as the feature amount extracted by the feature amount calculation unit 13. At that time, the feature quantity output unit 14 associates the feature quantity with the ID information of the fingerprint image corresponding to the feature quantity, and outputs it to the database 20. The feature amount output unit 14 may output the singular points (core singular points and delta singular points) of the fingerprint image to the database 20 in addition to the calculated feature amounts.

[Configuration and operation of input device]
Next, the configuration and operation of the input device 30 will be described.

  FIG. 20 is a diagram illustrating an example of the internal configuration of the input device 30. Referring to FIG. 20, the input device 30 includes a data input unit 31, a data output unit 32, and a storage unit 33.

  The data input unit 31 inputs image data related to a retained fingerprint and ID information of the retained fingerprint through an external storage device (USB memory or the like) or a network. In addition, the data input unit 31 accepts an operation by an inspector, generates an operation screen corresponding to the operation, and displays the operation screen on a display device (not shown) such as a liquid crystal display. That is, the data input unit 31 displays a screen for a user (operator) operation on the display device, and receives an operation for inputting a feature amount of a feature point by the user.

  The inspector finds singular points (core type singular points and delta type singular points) and feature points from the fingerprint image (mainly retained fingerprint images) on the operation screen displayed on the display device, and inputs information and feature amounts related to the positions. To do. For example, when inputting a feature point direction such as the first feature point direction or the second feature point direction, the inspector inputs the feature point direction by performing a drag operation along the input feature point direction. To do. Thereafter, when the drag operation by the inspector is completed, the data input unit 31 displays a pull-down menu as shown in FIG. 21, and the inspector performs “setting of the first feature point direction” or “second feature point direction”. In response to selecting “SET”, the feature point direction is input.

  When the input device 30 displays the feature point direction input by the inspector, the display mode of the first feature point direction and the second feature point direction are different from each other as shown in FIG. (For example, one has a bulge, the other does not have a bulge, uses a dotted line, or changes a display color). By providing such an operation interface to the inspector, the inspector can input the feature amount by an intuitive operation.

  The data input unit 31 stores the feature amount of each feature point input in this way in the storage unit 33. At that time, the data input unit 31 executes a process such as converting the feature point direction indicated by the drag operation into an angle shown in FIG. 7, and is the same as the feature quantity registered in the database 20 by the feature quantity generation device 10. Are stored in the storage unit 33. Here, the input device 30 includes the fingerprint image input unit 11, the core image extraction unit 12, the feature amount calculation unit 13, and the feature amount output unit 14 described above, and the feature amount and ID information of the image data related to the retained fingerprint. May be displayed on the display device. With such a configuration, the inspector only has to correct when there is an error in the feature amount related to the displayed residual fingerprint, and the processing time of the inspector can be reduced.

  When the data input operation by the inspector is completed, the data output unit 32 reads the feature amount stored in the storage unit 33 and outputs the feature amount to the collation device 40 together with the ID information of the retained fingerprint image.

[Configuration and operation of verification device]
Next, the configuration and operation of the verification device 40 will be described.

  FIG. 22 is a diagram illustrating an example of the internal configuration of the collation device 40. Referring to FIG. 22, the collation device 40 includes a feature amount input unit 41, a feature point collation unit 42, a collation result output unit 43, and a storage unit 44.

  The feature amount input unit 41 inputs the feature amount and ID information related to the fingerprint image on the search side (search side) in fingerprint collation from the input device 30. Further, the feature amount input unit 41 accesses the database 20 in response to a request from the feature point matching unit 42, and acquires the feature amount and ID information related to the fingerprint image to be searched (file side) in fingerprint matching. To do. The feature amount input unit 41 delivers the acquired feature amount and ID information (search side and file side feature amounts and ID information) to the feature point matching unit 42.

  The feature point matching unit 42 performs matching of the fingerprint image on the search side and the file side based on the feature amount related to the fingerprint (retained fingerprint) image on the search side and the feature amount related to the fingerprint (print fingerprint) image on the file side. And a collation score indicating the collation result is calculated. The feature point matching unit 42 performs the feature point matching process disclosed in “4 Fingerprint Matching” of Non-Patent Document 1 regarding the matching of the position and the first feature point direction when matching the two given feature quantities. Can be used.

  Here, the matching process using the second feature point direction by the feature point matching unit 42 will be mainly described with reference to FIG. 23 and subsequent drawings. Note that the fingerprint image shown on the left side of FIG. 23A is an image related to a deceased fingerprint, and the feature points and feature quantities shown in the figure are input by an inspector. Also, the fingerprint image shown on the right side of FIG. 23A is an image related to the person's imprinted fingerprint corresponding to the left fingerprint image, and the feature points and their feature quantities shown in FIG. Is calculated.

  The feature point matching unit 42 includes a plurality of feature points 401 to 405 illustrated on the left side of FIG. 23A and a plurality of feature points illustrated on the right side (ten feature points including feature points 411 to 415). And the correspondence between them is detected. For example, the correspondence between the feature point 401 on the left side (search side; left fingerprint side) and the feature point 411 on the right side (file side; fingerprint fingerprint side), the left side feature point 402 and the right side feature point 412 Corresponding relationships between them are detected.

  In calculating the correspondence between feature points, the feature point matching unit 42 detects the correspondence in consideration of the second feature point direction. More specifically, the feature point matching unit 42 detects the correspondence between feature points according to the matching rules (1) to (3) shown below. At that time, the feature point collating unit 42 performs alignment processing (positioning) of two fingerprint images and converts the feature point coordinates. Various methods can be employed to align the two fingerprint images. For example, two fingerprint images may be aligned using singular points (core singular points and delta singular points). If the two feature point coordinates on the search side and the file side after the coordinate conversion in the alignment processing are within a predetermined range, the feature point matching unit 42 determines that the two feature points are candidates for the feature points. deal with. After that, the feature point matching unit 42 compares the feature amounts (first and second feature point directions) of the paired feature point candidates, and performs final determination (paired feature point, unpaired feature point). .

  Collation Rule (1) When the positional relationship between two feature points is within a predetermined range, if both feature points do not have the second feature point direction, the first feature point of each feature point Correspondence is detected according to whether or not the feature point directions substantially match (determined as a feature point). That is, when the two singular point coordinates on the search side and the file side after coordinate conversion by the alignment process are within a predetermined range, it is determined that the two feature points are present at the same position. Thereafter, if the first feature point directions of the feature points coincide with each other, it is determined that the two feature points are the pair feature points.

  Collation rule (2) When the positional relationship between two feature points is within a predetermined range, one feature point has only the first feature point direction, and the other feature point has the first and second feature point directions. If so, the correspondence is detected according to the following conditions.

  (2-1) When the feature point type (branch point, end point) is the same, the first feature point direction of the feature point having the first feature point direction is the first and second feature points of the other feature point. A correspondence is detected if it substantially matches one of the directions.

  (2-2) When the feature point types are different, the feature point direction of the feature point having the first feature point direction and the feature point having the first and second feature point directions is any of the following relationships: If the condition is satisfied, a correspondence relationship is detected between the two feature points.

  The first relationship is a case where the first feature point direction of the feature point having the first feature point direction substantially coincides with one of the first and second feature point directions of the other feature point. . A correspondence relationship is detected between two feature points that satisfy the first relationship.

  The second relationship is substantially the same as the direction in which the first feature point direction of the feature point having the first feature point direction bisects the interior angle formed by the straight line formed by the first and second feature point directions of the other feature point. Is the case. In the following description, a direction that bisects an internal angle formed by a straight line by the first and second feature point directions is referred to as a first verification direction and is illustrated using a dotted line in the drawings. A correspondence relationship is also detected between two feature points that satisfy the second relationship. That is, if the feature point having the first feature point direction is a branch point, the first feature point direction of the branch point is compared with the first matching direction of the end points whose positional relationship is within a predetermined range. Thus, the correspondence between the two feature points is determined. Similarly, if the feature point having the first feature point direction is an end point, the first feature point direction of the end point is compared with the first matching direction of the branch point whose positional relationship is within a predetermined range. Thus, the correspondence between the two feature points is determined.

  The case where the feature point types are the same will be described in detail with reference to FIG. 24. The feature point 406 has a first feature point direction, and the feature point 416 has two feature point directions. In the case shown in FIG. 24A, since the first feature point direction of the feature point 406 matches the first feature point direction of the feature point 416, a correspondence relationship is detected between the two feature points. In the case shown in FIG. 24B, since the first feature point direction of the feature point 406 matches the second feature point direction of the feature point 416, a correspondence relationship is detected between the two feature points. On the other hand, in the case shown in FIG. 24C, the first feature point direction of the feature point 406 does not match either the first feature point direction or the second feature point direction of the feature point 416, and therefore corresponds to two feature points. It is determined that there is no relationship (correspondence is not detected).

  Next, the case where the feature point types are different will be specifically described with reference to FIG. Referring to FIG. 25, the feature point 407 having the first feature point direction is a branch point. A feature point 417 having the first and second feature point directions is an end point. Therefore, in the example of FIG. 25, the correspondence between the two feature points is based on whether the two feature points 407 and 417 have the first relationship (2-2) or the second relationship. Is determined. In the case of FIG. 25A, the first feature point direction of the feature point 407 and the first feature point direction of the feature point 417 substantially coincide with each other, and thus the first relationship is satisfied. Accordingly, a correspondence relationship is detected between the two feature points. In the case of FIG. 25B, the first feature point direction of the feature point 407 and the first matching direction of the feature point 417 (the direction dividing the first and second feature point directions) are substantially the same. Therefore, the second relation is satisfied. Accordingly, a correspondence relationship is detected between the two feature points. On the other hand, referring to FIG. 25C, the first feature point direction of the feature point 407 does not coincide with either the first feature point direction or the second feature point direction of the feature point 417, and is also the first matching direction. It does not match. That is, since the two feature points do not satisfy both of the first and second relationships, no correspondence is detected between the two feature points.

  Matching rule (3) When the positional relationship between two feature points is within a predetermined range and both feature points have the first and second feature point directions, the corresponding relationship is detected according to the following conditions: Is done.

  (3-1) When the positional relationship of the same kind of feature points is within a predetermined range and each of the two feature points has the first and second feature point directions, the following four feature point directions When one of the combinations substantially matches, a correspondence relationship is detected between the two feature points. The first combination is a combination of one first feature point direction and the other first feature point direction. The second combination is a combination of one first feature point direction and the other second feature point direction. The third combination is a combination of one second feature point direction and the other first feature point direction. The fourth combination is a combination including one second feature point direction and the other second feature point direction. When each of the same kind of feature points has the first and second feature point directions, whether or not the feature point directions substantially coincide with each other in at least one of the first to fourth combinations. Correspondence between two feature points is determined.

  (3-2) Two feature points have different types, and an interior angle formed by straight lines in two feature point directions of any one of the two feature points is a predetermined threshold (for example, 100 degrees) In the following cases, the first matching direction (direction that divides the inner angle between the first feature point direction and the second feature point direction) is calculated from each of the two feature points, and the first matching direction is also calculated. In consideration, the correspondence between the two feature points is detected. Specifically, two directions (feature point directions) in at least one of the nine combinations of the first feature point direction, the second feature point direction, and the first matching direction that each of the two feature points has. , The matching direction) substantially matches, a correspondence relationship is detected between the two feature points.

  (3-3) The types of two feature points are different, and an interior angle formed by a straight line by two feature point directions of any one of the two feature points is a predetermined threshold (the above threshold; In the case of exceeding 100 degrees), in addition to the first matching direction, a direction that bisects the external angle formed by a straight line formed by the two feature point directions of each of the two feature points (hereinafter referred to as the second matching direction; one point in the drawing) (Shown using a chain line) is added to the comparison target. That is, at least one of the 16 feature combinations of the first feature point direction, the second feature point direction, the first matching direction, and the second matching direction that each of the two feature points has has two directions. If they substantially match, a correspondence relationship is detected between the two feature points. As a case where an internal angle formed by a straight line in two feature point directions exceeds a predetermined threshold (for example, 100 degrees), a case where a feature point exists in the delta region is assumed. That is, when a feature point exists in the delta region, constant evaluation is given to the two feature points regardless of the values in the first and second feature point directions.

  With reference to FIGS. 26 and 27, the above collation rule (3) will be specifically described.

  For example, referring to FIG. 26, the feature point 408 and the feature point 418 are both branch points and have first and second feature point directions. Therefore, in the example of FIG. 26, the rule (3-1) is used for detecting the correspondence between two feature points. In the case shown in FIG. 26 (a), the first and second feature point directions of the feature point 408 and the first and second feature point directions of the feature point 418 coincide with each other, so that the two feature points correspond to each other. A relationship is detected. 26B, the first feature point direction of the feature point 408, the second feature point direction of the feature point 418, the second feature point direction of the feature point 408, and the first feature point of the feature point 418. Since the directions match each other, the correspondence between the two feature points is detected also in this case.

  Referring to FIG. 27, the feature point 409 and the feature point 419 have different relationship between the feature point types. Therefore, in the example of FIG. 27, the rule (3-2) or (3-3) is used to detect the correspondence between two feature points. In the case of FIG. 27A, since the interior angle formed by the straight lines in the first and second feature point directions of each feature point is equal to or less than a threshold value (for example, 100 degrees), the first matching direction is also taken into consideration. When at least one of the nine combinations of the three directions of the feature points substantially matches the two directions, it is determined that the two feature points have a correspondence relationship. In the case of FIG. 27A, since the second feature point direction of the feature point 409 substantially matches the first matching direction of the feature point 419, a correspondence relationship is detected between the two feature points. .

  In the case shown in FIG. 27 (b), since the internal angle formed by the straight lines of the first and second feature points of each feature point exceeds a threshold value (for example, 100 degrees), the first and second matching directions are also taken into consideration. In any case, it is determined that there is a correspondence between two feature points when at least one of the sixteen combinations including four directions of each feature point substantially matches. In the case of FIG. 27B, the first matching direction of the feature point 409 and the second matching direction of the feature point 419 are at least matched, so it is determined that there is a correspondence between the two feature points. The That is, as shown in FIG. 27B, even when a feature point direction that matches the first and second feature point directions of the two feature points cannot be found, the feature point exists in the delta region ( If the interior angle exceeds 100 degrees), it can be determined that there is a correspondence between the two feature points.

  When the matching rules (1) to (3) are applied to the two fingerprint images shown in FIG. 23A, five feature points are detected.

  Referring to FIG. 23A, a left feature point 402 and a right feature point 412 are feature points each having a second feature point direction. FIG. 23B is a diagram in which feature points 402 are extracted, and FIG. 23C is a diagram in which feature points 412 are extracted. Referring to FIGS. 23B and 23C, the relationship shown in FIG. 26B is recognized between the feature point 402 and the feature point 412, and the corresponding relationship is detected.

  In FIGS. 23B and 23C, if the second feature point direction is not taken into consideration, the first feature point directions of the two feature points are different from each other, and thus no correspondence is detected between the two feature points. Therefore, the collation accuracy of the two fingerprint images shown in FIG. However, the feature point matching unit 42 according to the first embodiment can detect that the two feature points are paired by calculating the correspondence between the feature points together with the second feature point direction. is there.

  Here, in the vicinity of the core region and the delta region, the position of the end point tends to change (elongate / shrink) easily due to pressure at the time of printing, dirt on the left fingerprint image, and the like. If the end point position changes in the vicinity of the core, the first feature point direction also changes. However, by introducing the second feature point direction, the influence of such end point position change can be mitigated. . That is, when the two feature points are within a predetermined range and the respective feature points have the first and second feature point directions according to the matching rule of (3) above, the two feature points include Correspondence is often detected. Alternatively, even if one of the first and second feature point directions changes due to a change in the end point position, it is considered rare that the two feature point directions change at the same time. It is considered that at least one of the feature point directions matches. In such a case, the deterioration of collation accuracy is suppressed to the minimum.

  In addition, in the vicinity of the core area or the delta area, a change in the feature point type (from a branch point to an end point or vice versa) may occur due to pressure at the time of stamping or dirt on a leftover fingerprint image. Even in such a case, degradation of matching accuracy can be suppressed by using the first and second feature point directions for matching.

  For example, the fingerprint image (core data) shown in FIG. 28 (a) is registered on the file side as an imprint fingerprint, and the person's retained fingerprint corresponding to the fingerprint image of FIG. 28 (a) is shown in FIGS. 28 (b) to (d). It is assumed that any one of In the case of FIGS. 28B to 28D, the branch point is changed to the end point due to the influence of noise or the like. Two feature point directions are calculated from the fingerprint image shown in FIG. In addition, the first feature point direction is calculated from the fingerprint images of FIGS. In the fingerprint image collation shown in FIGS. 28A and 28B, the feature point correspondence is detected in accordance with the collation rule (2-2). In this case, since the first feature point direction in FIG. 28A and the first feature point direction in FIG. 28B substantially match, it is determined that the two feature points have a correspondence relationship. Similarly, in the collation of FIG. 28A and FIG. 28C, the second feature point direction of FIG. 28A and the first feature point direction of FIG. It is determined that there is a correspondence relationship between the two feature points.

  The branch point in FIG. 28A may transition to the end point shown in FIG. In the fingerprint image matching shown in FIGS. 28A and 28D, the correspondence relationship between the feature points is detected according to the matching rule of (3-2) or (3-3). Since FIG. 28 illustrates feature points in the vicinity of the core region, the interior angle formed by straight lines in the first and second feature point directions is equal to or less than a threshold value (for example, 100 degrees). Therefore, the collation of the fingerprint images shown in FIGS. 28A and 28D follows the collation rule (3-2). In this case, the first matching direction shown in FIG. 28A and the first feature point direction shown in FIG. 28D substantially coincide with each other, so a correspondence relationship is detected between the two feature points.

  Further, for example, in the example of FIG. 18, when the end point 334 expands and the intersection point 372 and the end point 334 are combined, the fingerprint image shown in FIG. 18A is determined as a branch point of the delta region. Since the calculation method of the first feature point direction differs between the branch point and the end point, if only the first feature point direction is used for the feature point matching, the correspondence between the two feature points is calculated. Rare (not a feature point). However, even if the end point is expanded to change to a branch point, the accuracy of matching can be improved by introducing the second feature point direction.

  Consider collation of two feature points (end points and branch points) shown in FIGS. 29 (a) and 29 (b). In FIG. 29, each feature point has two feature point directions and exists in the vicinity of the delta region (because the internal angle by the first and second feature point directions exceeds a threshold value (for example, 100 degrees)), Follow the collation rule (3-3) above. The first and second collating directions are calculated from the two feature points, and when the match / mismatch of each direction is determined for 16 combinations, at least the first collating direction in FIG. 29A and FIG. ) Substantially coincide with the first feature point direction. Accordingly, a correspondence relationship is detected between the two feature points. In this way, since the positions of the two feature points are within a predetermined range, and the correspondence between the two feature points having the first and second feature point directions can be detected at least, the two feature points in the delta region can be detected. The feature point is often determined as a pair feature point. That is, by introducing the second feature point direction, the matching accuracy is improved.

  When the feature point matching unit 42 according to the first embodiment detects the correspondence between two feature points, the interior angle in the first and second feature point directions exceeds a predetermined threshold (for example, 100 degrees). In this case, the second matching direction is calculated, but the feature amount calculation unit 13 may calculate the direction. That is, when the internal angle of the first and second feature point directions exceeds a predetermined threshold (for example, 100 degrees), the feature amount calculation unit 13 determines the third feature point direction different from the first and second feature amounts. May be calculated as a feature amount that characterizes the feature point.

  The feature point matching unit 42 calculates a matching score based on the correspondence certainty for each of the feature points (5 in the case of FIG. 23) for which the correspondence is detected and the number of feature points for which the correspondence is not detected. To do.

  The feature point matching unit 42 changes the correspondence certainty factor according to various modes shown in FIGS. 24 to 27 when calculating the correspondence certainty factor regarding the feature point having the second feature point direction. The correspondence relationship may be differentiated.

  For example, referring to FIGS. 26 and 27, among the correspondence relationships shown in FIGS. 26 and 27, the correspondence relationship between the two feature points shown in FIG. 27 is the feature point shown in FIGS. 26 (a) and (b). It is considered weaker than the correspondence between the two. That is, in FIGS. 26 (a) and (b), at least one feature point direction matches, whereas the two feature points shown in FIGS. 27 (a) and 27 (b) match the feature point direction. This is because there is nothing. Therefore, the feature point matching unit 42 gives a low degree of confidence to the two feature points having the correspondence relationship as shown in FIG.

  On the other hand, in FIG. 26A, the first and second feature point directions of the two feature points coincide with each other. In such a case, the correspondence between the two feature points 407 and 417 is considered to be stronger than the correspondence between the two feature points shown in FIG. The two feature points shown in FIG. 26A are given higher correspondence certainty than the correspondence between the two feature points shown in FIG.

  The feature point matching unit 42 delivers the matching score to the matching result output unit 43 as a matching result. Alternatively, the feature point matching unit 42 may deliver a list of feature points in which the correspondence is detected in addition to the matching score to the matching result output unit 43 as a corresponding feature point list. In the corresponding feature point list, the coordinates of the feature points on the search side (retained fingerprint side) where the correspondence relationship is detected and the coordinates of the feature points on the file side (printed fingerprint side) that are paired with the feature points are included. Recorded in association.

  The collation result output unit 43 associates the ID information of the residual fingerprint image that is the collation target, the ID information of the fingerprint image for which the collation score is calculated, and the collation score, and sets the collation result as an external device (FIG. (See FIG. 30).

  As a method for dealing with the problem that the feature point direction cannot be stably calculated in the core region or the delta region, it is conceivable to relax the collation criteria in the core region or the delta region. That is, when matching the feature points of the core region and the delta region, the allowable range regarding the difference in the feature point direction is expanded. However, such an increase in the allowable range causes a new problem that it involves a phenomenon of increasing the matching score of a fingerprint that is not a pair (non-fingerprint). On the other hand, in the first embodiment, since the allowable range is not expanded as described above, it is possible to suppress the problem of increasing the verification score of the non-fingerprint.

  The overall operation of the fingerprint collation system according to the first embodiment is summarized as shown in the sequence diagram of FIG.

  The feature quantity generation device 10 inputs a fingerprint image (step S01), and extracts a core line image from the fingerprint image (step S02). Thereafter, the feature quantity generation device 10 calculates a feature quantity from the core line image (step S03). At that time, the feature quantity generation device 10 also calculates the second feature point direction from the feature points of the region where the feature point direction cannot be stably calculated, such as the core region and the delta region. The feature quantity generation device 10 registers the calculated feature quantity in the database 20 (step S04).

  The input device 30 accepts an operation by an inspector and inputs a feature amount of the fingerprint image (step S11). The input device 30 outputs the input feature quantity toward the collation device 40 (step S12).

  The collation device 40 acquires the feature quantity output from the input device 30 and the feature quantity registered in the database 20 (step S21), and a fingerprint image (for example, a survivor) corresponding to the feature quantity output from the input device 30. A fingerprint image) matching process is performed (step S22). At that time, the collation device 40 performs collation of the fingerprint image in consideration of the second feature point direction attached to each feature point. The collation apparatus 40 outputs a collation result as a collation score (step S23).

  As described above, the feature quantity generation device 10 according to the first embodiment calculates the second feature point direction as necessary in addition to the first feature point direction. Considering this second feature point direction, the matching accuracy is improved by detecting the correspondence between the two feature points and calculating the matching score. By introducing the second feature point direction, for example, even if the first feature point direction differs between two feature points, if the first feature point direction of one and the second feature point direction of the other match, This is because one feature point can be operated as a pair of feature points.

[Second Embodiment]
Next, a second embodiment will be described in detail with reference to the drawings.

  In the first embodiment, it is assumed that feature points exist in the core region and the delta region, but rarely feature points may not exist in these regions. Specifically, when the core region is composed of simple loop-shaped core hoof lines, there are no feature points. For example, when the main region of the retained fingerprint is the core region, the retained fingerprint High matching accuracy cannot be expected.

  The core hoof line is a loop-like (hoof-like) ridge (core line) located on the innermost side in the fingerprint image, and has a feature point near its apex (for example, within one ridge interval). It means a core wire that does not exist. For example, referring to FIG. 32A, the curve 501 is a core hoof but the curve 502 is not a core hoof. This is because the curve 502 is not a ridge located on the innermost side of the fingerprint image. Alternatively, the curves 503 and 504 shown in FIG. 32B are not the core hoof lines. This is because these curves are not ridges formed in a loop shape.

  Even in the delta region, there may be no feature point near the center of the delta. For example, when the main region of the residual fingerprint is the delta region, high accuracy of fingerprint verification using such a small region fingerprint is expected. Can not. For example, referring to FIG. 33 (a), a protruding ridge is recognized in the vicinity of the delta, but this protruding ridge does not have a sufficient length (for example, one ridge interval), so it is removed as noise. And is not treated as a core wire. Referring to FIG. 33 (b), there is a dot-shaped ridge, but this ridge does not have a sufficient length, so it is removed as noise and not treated as a core.

  Furthermore, fingerprint images corresponding to the core area and delta area are vulnerable to noise, and feature point transitions easily (end points transition to branch points or vice versa), or new end points occur due to ridge breaks. Or For example, if there is no feature point in the core area on the side of the fingerprint fingerprint but there is a feature point on the left side of the fingerprint due to the influence of noise or the like, two fingerprints are considered even if the direction of the second feature point is considered. Image matching accuracy is degraded. In the first place, there is no feature point on the fingerprint side, and there is no matching target corresponding to the feature point on the left fingerprint side.

  In view of the situation as described above, in the second embodiment, a fingerprint verification system that improves fingerprint verification accuracy regardless of the presence or absence of feature points in the core region or the delta region will be described.

  FIG. 34 is a diagram illustrating an example of an internal configuration of the feature quantity generation device 10a according to the second embodiment. The difference from the feature value generation apparatus 10 according to the first embodiment is that a core wire adding unit 16 is added. Therefore, details of the core wire adding unit 16 will be described, and description of each functional block such as the feature amount calculating unit 13 will be omitted.

  The core line adding unit 16 is means for detecting a ridge or area having a predetermined characteristic and adding the core line to the detected ridge or area. More specifically, the core wire adding unit 16 detects a core wire or a region where a core wire needs to be added, and executes a predetermined core wire adding process when the region is detected.

  For example, the core line adding unit 16 detects the core hoof line from the core line image extracted by the core line image extracting unit 12. Specifically, the core line adding unit 16 extracts a core line that is the innermost loop-shaped core line in the fingerprint image and has no feature point near its vertex. Furthermore, the core wire adding unit 16 adds a pseudo-protruding core wire to the vertex of the detected core hoof line.

  It is assumed that a fingerprint image having an area as shown in FIGS. 35A and 35B is input to the feature quantity generation device 10a. In this case, when only the core line image extracted by the core line image extracting unit 12 is extracted, FIG. 35C is obtained. The core wire adding unit 16 detects the core hoof line 601 from FIG. 35 (c) and adds a short protruding core line (hereinafter referred to as a protruding core line) 611 to the apex of the core hoof line 601 (FIG. 35 ( d) Core line in the area surrounded by a circle). The direction of the pseudo-projection core line added by the core wire addition unit 16 is opposite to the direction of the core hoof line (the direction from the apex of the core hoof line 601 to the center of the core hoof line) in FIG. And the length is preferably about half of the ridge interval.

  Further, the core wire adding unit 16 uses, for example, three core wires that form a delta (triangular) shape from the core image extracted by the core image extracting unit 12 by using a method such as pattern matching, and the three core wires. An area to be included (delta area) is detected. Furthermore, the core line adding unit 16 adds a pseudo dot-shaped core line (hereinafter referred to as a delta dot core line) when there is no feature point near the center of the detected delta. For example, it is assumed that a fingerprint image having an area as shown in FIGS. 36A and 36B is input to the feature quantity generation device 10a. In this case, the core wire adding unit 16 detects the core wires 602 to 604 as core wires forming a delta.

  Next, the core wire adding unit 16 adds a delta dot core wire 612 near the center of the core wires 602 to 604 (region surrounded by a circle in FIG. 36C). In addition, it is preferable that the direction of the delta dot core line added by the core line addition unit 16 is horizontal with respect to the fingertip direction (the direction of the fingertip), and the length of the delta dot core line is about half of the ridge interval.

  By the processing of the core line adding unit 16, two feature points are generated for both the core region and the delta region. Specifically, referring to FIG. 35 (d), a branch point composed of the core hoof line 601 and the projection core wire 611 and an end point of the projection core wire 611 are generated. In addition, referring to FIG. 36C, both ends of the delta dot core wire 612 are generated as end points.

  When the processing by the core line adding unit 16 is completed, the feature amount calculating unit 13 extracts feature points and features (position, first and second features) by the same method as described in the first embodiment. (Point direction) is calculated.

  Here, the input device 30 described in the first embodiment also needs to correspond to an operation (operation by an inspector) for adding a core wire to the core region or the delta region.

  When an inspector adds a protruding core wire in the core region, the expert performs a drag operation in the core direction (the direction of the core hoof line) at the apex of the core hoof line. Thereafter, the input device 30a according to the second embodiment displays a menu (pull-down menu) as shown in FIG. The inspector selects “Add Projection Core Line” from the displayed pull-down menu to end the operation of adding the projection core line to the core hoof line. It should be noted that when the core line of the core hoof is added, the input device 30a is configured so that data can be input using the apex of the core hoof as a core (core singular point), thereby reducing the input man-hours for the inspector (operator). .

  The expert can also input the delta dot core wire using the input device 30a. When adding the delta dot core line to the delta area, the inspector performs a drag operation in the first delta direction from the delta position (delta center). The delta direction means three directions emitted from the delta. FIG. 38A shows an example in the delta direction.

  When the inspector finishes the drag operation for inputting the first delta direction, a pull-down menu as shown in FIG. 38B is displayed from the input device 30a. When the inspector selects “Add Delta Dot Core” from the displayed pull-down menu, the input device 30a instructs the inspector to input the remaining two delta directions. The input device 30a adds the delta dot core line to the center of the delta at the end of the operation for inputting the third delta direction (end of the drag operation for inputting the third delta direction). Alternatively, an instruction to input the direction of the delta dot core line may be given to the inspector.

  In addition, when the delta dot core wire is added, the input device 30a is configured so that the area where the operation is performed can be input as a delta (delta type singular point), thereby reducing the input man-hours for an expert (operator). .

  The collation device 40 performs a fingerprint image collation process using the feature amount generated by the feature amount generation device 10a and the feature amount input by the inspector using the input device 30a, and generates and outputs a collation result. To do.

  Next, the effect of adding a protruding core wire to the core region and a delta dot core wire to the delta region will be described.

  As described above, the image near the apex of the core hoof line changes due to pressure, dripping, dirt, etc. at the time of imprinting. As a result, the simple core hoof line is disconnected or connected to the outer ridge. May end up. Specifically, the core hoof line as shown in FIG. 39 (a) is originally broken along the way as shown in FIG. 39 (b), or the outer ridge line as shown in FIG. 39 (c). Sometimes connect.

  When transitions (changes) as shown in FIGS. 39B and 39C occur, two feature points in the same direction are generated. In this case, if the fingerprint image is the fingerprint image (core image) shown in FIG. 39 (a) and the retained fingerprint image is a fingerprint image as shown in FIG. 39 (b) or (c), the features in the core region The number of points will vary greatly. Then, two feature points are extracted from the core area of the left fingerprint, but there are no feature points corresponding to the two feature points in the fingerprint, and the two feature points are treated as unpaired feature points. As a result, the collation score in the fingerprint image related to the imprint fingerprint and the deceased fingerprint decreases.

  On the other hand, the feature quantity generation device 10a according to the second embodiment creates feature points corresponding to two feature points that are likely to occur due to the influence of dirt or the like by adding a projection core wire to the apex of the core hoof line. . As a result, a situation in which no feature point exists in the other fingerprint image does not occur on either the search side or the file side, and a decrease in matching score due to the absence of the corresponding feature point is suppressed. .

  In addition, as described in the first embodiment, since it is difficult to stably calculate the feature point direction in the core region, the second feature point direction that complements the first feature point direction is calculated. The The two feature points that have two feature point directions and are close to each other can be detected in correspondence (the matching rule (3) described in the first embodiment), so that a protrusion core is added. It is possible to pair two feature points of each fingerprint image in which the feature points are created due to the influence of the fingerprint image and dirt and the like, and deterioration of collation accuracy is suppressed.

  In addition, as described above, feature points may not be extracted from the delta region. Alternatively, as in the core region, the image near the delta changes due to the influence of pressure, wrinkles, and dirt during printing, and as a result, a short ridge (dot or protrusion) may be extracted near the delta. In other words, feature point extraction is unstable in the delta region, and there is a tendency that a feature point does not exist in the fingerprint image on the search side and a feature point exists in the fingerprint image on the file side. For example, feature points are not extracted from the fingerprint image on the file side as shown in FIG. 40A, and feature points due to short ridges are shown in the fingerprint image on the search side as shown in FIG. May be extracted.

  In such a case, there are no two corresponding feature points between the two fingerprint images, and the two feature points in FIG. 40B are treated as unpaired feature points, which causes a decrease in the matching score. However, in the second embodiment, for example, two feature points are created by adding a delta dot core line to the fingerprint image shown in FIG. Therefore, the feature points of the two fingerprint images can be easily paired, and the matching score increases.

  As described above, in the second embodiment, when there are no feature points in the core region or the delta region, a process of adding a core line to these regions is performed. Since the added feature line calculates the second feature point direction described in the first embodiment, the correspondence between the feature points existing in the core region and the delta region can be detected more reliably. As a result, fingerprint image matching accuracy is improved. In the second embodiment, the description has been made on the assumption that the second feature point direction described in the first embodiment is used. However, the second feature point direction may not be used. That is, in the second embodiment, if there are no feature points in the core region or the delta region, a protruding core wire or a delta dot core wire is added so that the feature points are generated in these regions. By adding these core lines, there are no feature points in the core region and delta region of one fingerprint image (core line image), and there are feature points in the core region and delta region of the other fingerprint image. The situation can be avoided, and the deterioration of the collation accuracy can be suppressed (the collation accuracy is improved).

  Note that the configuration of the fingerprint collation system described in the first and second embodiments (FIG. 2) is an example, and is not intended to limit the configuration of the system. For example, all or some of the various functions of the feature quantity generation device 10, the input device 30, and the collation device 40 may be incorporated in another device. Alternatively, various functions of the feature quantity generation device 10, the input device 30, and the collation device 40 may be incorporated in one device. When the functions of the feature quantity generation device 10 and the like are realized by one device, an image processing device 50 shown in FIG. 41 can be used. Note that functional blocks such as the fingerprint image input unit 11 shown in FIG. 41 are the same as the functional blocks described in the first and second embodiments, and thus description thereof is omitted. However, since a means for accessing the database 20 is necessary, the image processing apparatus 50 includes a database access unit 51. The feature amount calculation unit 13 registers the generated feature amount in the database 20 via the database access unit 51. The feature point matching unit 42 acquires the feature amount registered in the database 20 via the database access unit 51. Of course, the database 20 may be constructed in the storage unit 15.

  In the first and second embodiments described above, the configuration and operation of the feature quantity generation device 10 and the like have been described using a fingerprint image as an image in which a curved stripe pattern is formed by ridges. It is not limited. For example, the feature value generation apparatus 10 may calculate a feature value from an image related to a palm print or the like.

  The feature amount calculation processing described in the first and second embodiments is to calculate the second feature point direction in addition to the first feature point direction. It is also possible to calculate. That is, it is also possible to calculate the first to third feature point directions as feature amounts that characterize one feature point. For example, referring to FIG. 14, since the difference between the three interior angles is small at the branch point in the delta region, the interior angle a3 (maximum interior angle) is bisected rather than the feature point direction by the minimum interior angle a1 and the second minimum interior angle a2. It is also assumed that the feature point direction is calculated as the first or second feature point direction. In such a case, since the collation score is expected to deteriorate, the direction that divides the maximum inner angle a3 into two can be calculated as the third feature point direction and used for the feature point collation.

  However, the inconvenience as described above can be solved by a response at the time of collation processing. For example, the first feature point direction 304 and the second feature point direction 315 shown in FIG. 14 are extracted on the search side (remaining fingerprint side), and the direction that divides the maximum inner angle a3 into two on the file side (imprint fingerprint side). Consider a case where the direction is registered as the first feature point direction. In such a case, in the feature points in the delta region, the first feature point direction 304 and the second feature point direction are considered in consideration that the difference between the first feature point direction and the second feature point direction is about 120 degrees. Of the two angles formed by 315, the direction that bisects the larger angle may be calculated as the third feature point direction on the search side. That is, by comparing the third feature point direction on the search side and the first feature point direction on the file side, it is possible to suppress degradation of matching accuracy without introducing the third feature point direction.

  In the first and second embodiments, the direction that divides the interior angle into two is determined as the feature point direction when calculating the first and second feature point directions, but the values for dividing the angle are other values. May be. The ratio of dividing the angle when calculating the feature point direction is not limited to a ratio of 1: 1.

  Further, the processing performed by each unit of the feature quantity generation device 10, the input device 30, and the collation device 40 is performed on the hardware mounted on the computer mounted on these devices (the feature quantity generation device 10, the input device 30, and the collation device 40). Can be realized by a computer program for executing the above-described processes. There may be a means for executing the function performed by the feature amount calculation unit 13 or the like by some kind of hardware and / or software. That is, the processing performed by each unit of the feature quantity generation device 10, the input device 30, and the collation device 40 is performed by a circuit such as a CPU (Central Processing Unit) or an ASIC (Application Specific Integrated Circuit) as shown in FIGS. It may be realized. There may be one or more circuits. The plurality of circuits may cooperate in the same device or may exist in different devices. The feature quantity generation device 10a is the same as the feature quantity generation device 10 of FIG.

  Furthermore, by installing the above-described computer program in the storage unit of the computer, the computer can function as an image processing apparatus. Furthermore, by causing a computer to execute the above-described computer program, it is possible to cause the computer to execute an image processing method related to feature amount calculation or the like. The program can be downloaded through a network or updated using a storage medium storing the program.

  Moreover, in the flowchart used by the above-mentioned description, although several process (process) is described in order, the execution order of the process performed by each embodiment is not restrict | limited to the description order. In each embodiment, the order of the illustrated steps can be changed within a range that does not hinder the contents, for example, the processes are executed in parallel.

A part or all of the above embodiments can be described as follows, but is not limited to the following.
[Form 1]
This is the same as the image processing apparatus according to the first viewpoint described above.
[Form 2]
The core wire addition part is
A loop-like ridge located on the innermost side of the core image, wherein a core hoof line having no feature point within a predetermined range from the apex is detected, and a protrusion-like shape is formed at a substantial apex of the core hoist line. The image processing apparatus according to aspect 1, wherein a core wire is added.
[Form 3]
The core wire addition part is
The image processing apparatus according to the first or second aspect, wherein a delta area is detected from the core line image, and a core line is added to the center position when a core line does not exist at a substantial center position of the detected delta area.
[Form 4]
A feature quantity calculating unit for calculating first and second feature point directions characterizing the feature points at both ends of the added core line; and
A feature amount output unit that outputs the calculated first and second feature point directions as feature amounts of the extracted feature points;
An image processing apparatus according to any one of Forms 1 to 3, further comprising:
[Form 5]
A screen for user operation is displayed on a display device, and further includes a data input unit that receives an operation of inputting a feature amount of a feature point by the user,
In the first to fourth aspects, when the data input unit displays the first and second feature point directions on the display device, the display modes of the first and second feature point directions are different from each other. The image processing apparatus according to any one of the above.
[Form 6]
This is the same as the image processing method according to the second viewpoint described above.
[Form 7]
It is as the program which concerns on the above-mentioned 3rd viewpoint.
Form 6 and form 7 can be developed like form 2 to form 5 in the same manner as form 1.

  Each disclosure of the cited patent documents and the like cited above 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.

10, 10a Feature amount generation device 11 Fingerprint image input unit 12, 102 Core line image extraction unit 13 Feature amount calculation unit 14 Feature amount output unit 15, 33, 44 Storage unit 16, 103 Core line addition unit 20, 20a Database 30, 30a Input Device 31 Data input unit 32 Data output unit 40 Collation device 41 Feature amount input unit 42 Feature point collation unit 43 Collation result output unit 50, 100 Image processing unit 51 Database access unit 101 Image input units 201-205, 321 Branch point 211- 213, 331 to 334 End points 221 to 243 Feature point directions 231 to 237, 341 to 349 End points 301 to 304, 351 First feature point directions 311 to 315, 352, 353 Second feature point directions 361 and 362 Virtual lines 371, 372 Intersection points 401 to 409, 411 to 419 Characteristic points 501 to 504 Curve 60 1 Core hoof wire 602 to 604 Core wire 611 Protruding core wire 612 Delta dot core wire 701 Core region 702 Delta region 703 region

Claims (7)

An image input unit for inputting an image in which a curved stripe pattern is formed by ridges;
A core image extraction unit that generates a core image obtained by extracting a core from the image;
A core line adding unit for detecting a ridge or area having a predetermined characteristic from the core line image and adding a core line to the detected ridge or area;
An image processing apparatus comprising: The core wire addition part is
A loop-like ridge located on the innermost side of the core image, wherein a core hoof line having no feature point within a predetermined range from the apex is detected, and a protrusion-like shape is formed at a substantial apex of the core hoist line. The image processing apparatus according to claim 1, wherein a core wire is added. The core wire addition part is
The image processing apparatus according to claim 1, wherein a delta region is detected from the core line image, and a core line is added to the center position when no core line exists at a substantial center position of the detected delta region. . A feature quantity calculating unit for calculating first and second feature point directions characterizing the feature points at both ends of the added core line; and
A feature amount output unit that outputs the calculated first and second feature point directions as feature amounts of the extracted feature points;
The image processing apparatus according to claim 1, further comprising: A screen for user operation is displayed on a display device, and further includes a data input unit that receives an operation of inputting a feature amount of a feature point by the user,
The said data input part shall mutually differ the display mode of the said 1st and 2nd feature point direction, when displaying the said 1st and 2nd feature point direction on the said display apparatus. The image processing apparatus according to any one of the above. Inputting an image in which a curved stripe pattern is formed by ridges;
Generating a core image obtained by extracting a core from the image;
Detecting a ridge or region having a predetermined characteristic from the core image, and adding a core to the detected ridge or region;
Including an image processing method. Processing to input an image in which a curved stripe pattern is formed by ridges;
A process of generating a core image obtained by extracting a core line from the image;
Detecting a ridge or region having a predetermined characteristic from the core image, and adding a core to the detected ridge or region;
For causing a computer mounted on the image processing apparatus to execute the program.