JP5279007B2 - Verification system, verification method, program, and recording medium - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To propose a verification system for preventing the information leakage of biological information in biometrics authentication using a portable communication terminal equipped with an image shooting function, and for achieving the reduction of the labor of authentication processing while securing the precision of authentication. <P>SOLUTION: For example, as for a palmistry service, a portable communication terminal 3<SB>l</SB>equipped with an image photographing function creates characteristics data in one direction based on photographed biological information so that the user of the portable communication terminal 3<SB>l</SB>can make the biological information itself secret. Also, it is possible for a server 5 to reduce calculation amounts in the portable communication terminal 3<SB>l</SB>by using an approximation verification processing, and to achieve precision assurance in collecting an image, and it is possible for a user to happily use the service by freely collecting an image without worrying about the precision of the image. Furthermore, it is possible even for a portable communication terminal whose calculation resources are restricted to extract characteristic points with high precision by using lifting Haar-like characteristics. <P>COPYRIGHT: (C)2010,JPO&amp;INPIT

Description

  The present invention relates to a collation system, a collation method, a program, and a recording medium, and in particular, collation based on a portable communication terminal having an image photographing function and image information obtained by photographing with the image photographing function of the portable communication terminal. The present invention relates to a verification system including a server that performs processing.

  Opportunities for identity verification are increasing in situations where high security is required, such as finance and immigration. In order to verify the identity, biometric authentication using information related to a human living body has attracted particular attention. There are two types of biometric information used for authentication: physical characteristics and behavioral characteristics. The former corresponds to a fingerprint, a palm shape, a face, an iris, a vein, and the like, and the latter is typically a voice print, a signature, and the like. Biometric authentication is more convenient than conventional authentication methods because it does not require possession of a portable device such as an IC card or remembering a password, and can reduce the burden on the user. In addition, since a unique characteristic of biological information such as uniqueness in which all people have different information is used, a system that is effective against attacks such as impersonation can be constructed.

  However, it is difficult to change the living body itself that is the source of the biological information. Therefore, once biometric information leaks, the same biometric information cannot be used safely. From the viewpoint of fault tolerance, which is one of the basic concepts of authentication security, this can be said to be more vulnerable than authentication using simple secret information. Among these problems, in particular, leakage of biometric information registered in advance on the certifier side such as a service provider is registered by converting the biometric information using a one-way function or appropriate encryption. Solutions have been proposed that allow the cancellation of information. Biometric authentication based on this idea is called revocable (see Non-Patent Document 1).

N. K. Ratha, 3 others, “Generating Cancelable Fingerprint Templates”, IEEE Transactions on pattern analysis and machine intelligence, vol.29, no.4, pp.561-572

  However, if biometric authentication is introduced into the ubiquitous system in the future, the possibility of information leakage is expected to increase.

  Furthermore, the conventional biometric authentication is based on the premise that biometric information collection conditions are within a certain range. However, for example, when biometric authentication is realized by capturing biometric information with a portable device having an image capturing function such as a digital camera or a mobile phone, the biometric information is subject to various imaging conditions (distance, light quantity, etc.). Taken at. For this reason, the amount of information processing increases, and it becomes difficult to implement in an environment in which computing resources such as portable devices are limited. On the other hand, even if the amount of information is simply reduced to simplify the process, the accuracy of the authentication process will deteriorate. This is a problem that generally exists in the verification process using biometric information.

  Therefore, the invention of the present application prevents the leakage of biometric information and guarantees the accuracy of authentication in biometric authentication using a portable communication terminal equipped with an image capturing function, which can be applied to, for example, a fortune telling service. The purpose is to propose a collation system that can be realized while reducing the burden of authentication processing.

A first aspect of the present invention is a collation system including a portable communication terminal having an image photographing function and a server that performs a collation process based on image information obtained by photographing with the image photographing function of the portable communication terminal. The mobile communication terminal includes an image capturing unit that acquires biometric information as image data, an image processing unit that generates feature data in one direction based on image information obtained by capturing, and the feature data to the server. Characteristic data transmitting means for transmitting, and the server includes characteristic data receiving means for receiving the characteristic data, and approximate matching means for performing an approximate matching process on the feature data and data stored in a database. .

The second aspect of the present invention is the first aspect of the verification system, the image data is palmprint image data, the image processing means extracts a feature point from an image represented by the palm print image data A feature point extraction unit; and a character string generation unit configured to generate a character string from the extracted feature point, wherein the feature data transmission unit transmits the character string as the feature data to the server. The approximate matching means performs an approximate character string matching process.

According to a third aspect of the present invention , there is provided a collation system according to the second aspect , wherein the character string generation means has a point of the image represented by the palmprint image data as a center point, and is clockwise or counterclockwise from the center point. The feature points are arranged around the character string.

According to a fourth aspect of the present invention , there is provided a collation system according to the second or third aspect , wherein the image processing means uses a plurality of feature images obtained by converting the palmprint image data using Haar-like features. Then, a new feature image is generated by weighting and combining with the lifting parameters.

A fifth aspect of the present invention is the collation system according to any one of the second to fourth aspects , wherein the data stored in the database is information that can restore the biological information of the user of the mobile communication terminal. Is based on a fortune telling service, the server is a collation result determining unit for determining a collation result of the approximate collating unit, and a determination result of transmitting a determination result by the collation result determining unit to the mobile communication terminal The mobile communication terminal includes a determination result receiving unit that receives the determination result, and a determination result display control unit that causes the display unit to display the determination result.

A sixth aspect of the present invention is a collation in a collation system comprising a portable communication terminal having an image photographing function and a server that performs collation processing based on image information obtained by photographing with the image photographing function of the portable communication terminal. A method in which image processing means of the mobile communication terminal generates feature data in one direction based on biological information acquired as image data, and the feature data transmitting means of the mobile communication terminal comprises the feature A step of transmitting data to the server; a step of receiving characteristic data of the server; a step of receiving the characteristic data; and an approximate matching unit of the server approximating the characteristic data and data stored in a database. Performing a verification process.

A seventh aspect of the present invention is a program for causing a computer to function as the image processing means according to any one of claims 1 to 5.

An eighth aspect of the present invention is a recording medium on which a program according to the seventh aspect is recorded.

  Note that the image capturing function provided in the mobile communication terminal is not limited to capturing biometric information, but is captured under various imaging conditions. For example, an image related to biometric information such as a palm print image. May be obtained.

In the first aspect , the image processing means may generate feature data using a one-way function.

Further, in the second aspect, the feature point extracting means may detect an edge and extract a feature point from the detected edge. Also, the feature of the palm wrinkle is within the range of the palm outline. As the point, at least one of an intersection, an end point, and a branch point may be extracted.

Further, in the second aspect , the character string generating means is based on a value obtained by quantizing the angle of the palm wrinkle line with respect to the horizontal right direction in each feature point and a value indicating the type of the feature point. After the generated label is attached to the coordinates of the feature points, each of the feature points is represented as a labeled coordinate, and then each of the labeled coordinates is defined according to a rule that each of the feature points is arranged clockwise from the center point. The set of feature points may be represented by a character string generated by sequentially combining.

Further, in the fourth aspect , the lifting parameters are determined so that the features in the class are close and the features between the classes are separated in a class configured by classifying each of the Haar-like features. May be.

  According to the present invention, biometric information is generated by generating feature data in one direction based on an image related to biometric information captured by a mobile communication terminal having an image capturing function, thereby realizing cancelable biometric authentication. It becomes possible to conceal itself. As a result, the user of the mobile communication terminal can manage the captured biometric information in his mobile communication terminal, and can use the service without worrying about information leakage from the server.

  In particular, in conventional biometric authentication, the server has been notified at least once about information that can restore the user's biometric information. However, for example, if it is a palm fortune-telling service according to the invention according to claim 5, the user of the mobile communication terminal does not need to notify the server of his / her biometric information as recoverable information, and information leakage from the server is prevented. You can use the service with peace of mind without worrying.

  In the present invention, feature point extraction / conversion processing is transferred to the mobile communication terminal in order to realize cancelable authentication. In general, ambiguity is included in the collection of images of biological information (for example, misalignment, enlargement / reduction, rotation, noise mixing, missing, extraction error, distortion, etc.). The amount of required feature point extraction / conversion processing varies depending on the accuracy at the time of image collection. Mobile communication terminals have limited computational resources, and it is important to reduce this processing. In the present invention, a gentle matching process is performed by the approximate matching process in the server. By using approximate collation processing instead of collation processing, the calculation amount is changed from linear to square, but the server has abundant calculation resources and can be realized. Thereby, the calculation amount in the mobile communication terminal can be reduced.

  Conventionally, accuracy guarantee at the time of image acquisition has been performed by fixing a living body or reducing noise using a dedicated scanner. However, in a mobile communication terminal, it is difficult to realize a dedicated scanner. As in the invention of the present application, by realizing the accuracy guarantee at the time of image collection by the approximate matching of the server, the user of the portable communication terminal can freely collect the image without worrying about the accuracy of the image and have fun. You can use the service.

  Further, as in the invention according to claim 2, for example, the type of the set feature point is used by using the matching with the one-dimensional character string information, instead of the matching with the two-dimensional information using the image as it is. -The amount of calculation can be further reduced by performing approximate character string matching that allows insertion, deletion, replacement, replacement, etc. based on the order.

  Furthermore, as described in the invention according to claim 4, by using the Haar-like feature, a feature point can be extracted with high accuracy even in a mobile communication terminal with limited calculation resources.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. The present invention is not limited to these examples.

  FIG. 1 is a block diagram showing a configuration and processing flow of a collation system 1 which is one embodiment of the present invention.

In FIG. 1, the verification system 1 includes a plurality of mobile communication terminals 3 _l (l = 1,..., N) and a server 5. The mobile communication terminal 3 _l captures the biometric information 7 (for example, the palm of the user of the mobile communication terminal 3 _l ) and acquires image information, and the image information acquired by the image acquisition unit 11 Are provided with an image processing unit 13 that performs image processing and generates feature data, and a feature data transmission unit 15 that transmits the feature data to the server 5. The server 5 also includes a feature data receiving unit 31 that receives feature data, a database 33, an approximate matching unit 35 that performs approximate matching between the feature data and information stored in the database 33, and a collation that determines an approximate matching result. a result determining section 37, a judging result transmitting unit 39 for transmitting the verification result of the determination to the mobile communication terminal 3 _l. The mobile communication terminal 3 _l, the judgment result and the determination result receiving unit 17 for receiving, and a display unit 21, the determination result display control unit that performs display control processing based on the determination result to the display screen of the display unit 21 19 is provided.

The processing performed in the mobile communication terminal 3 _l and the server 5, each specifically described.

First, a description will be given of a process performed by the mobile communication terminal 3 _l. Imaging function provided in the portable communication terminal 3 _l captures biometric information (e.g., fingerprint, hand geometry, etc.) in a variety of imaging conditions, on the basis of the captured image, the image information in the image acquisition unit 11 ( For example, a palm print image) is acquired.

Hereinafter, a case of a palm print image will be described as an example. Mobile communication terminal 3 _l imaging function provided in, for example, comprises assumes the general standard cameras such as digital cameras and camera-equipped mobile phone, a kind of dedicated scanners for biological information acquisition It is not a thing. For this reason, the image information acquired by the image acquisition unit 11 includes a lot of ambiguity (for example, misalignment, enlargement / reduction, rotation, noise mixing, missing, extraction error, distortion, etc.).

  The image processing unit 13 performs feature amount extraction / conversion processing on the palm print image. Below, the process which extracts the feature point called a minutia and converts the extracted minutia into a character string is demonstrated concretely. The image processing unit 13 includes a feature point extraction unit 23 that extracts an edge from a palmprint image and extracts a feature point from the edge information, and a character string generation unit 25 that generates a character string from the feature point information.

  Specifically, the minutiae is a characteristic point such as an end point indicating a break in the palm of the palm or a branch point which is a bifurcation. High accuracy is obtained by using the minutiae in identifying palm prints.

  FIG. 2A is a diagram for explaining intersections / end points / branch points extracted as minutiars in the first embodiment, and FIG. 2B is a diagram showing declination angles at the respective points (B, C, D).

  In this embodiment, extraction of the maneuver is applied to the palm image, and the minutia is an intersection where two wrinkle lines intersect, an end point where one wrinkle line is interrupted, and 1 Consider three types of branching points, which are portions where another one of the cocoon lines branches from the middle of the cocoon line (see FIG. 2A).

  In this embodiment, different types of minutiae are assigned different labels as follows depending on the angle of deviation of the lines constituting the minutia with respect to the horizontal direction. That is, with respect to the intersection, for the two wrinkle lines that intersect, the declination angle (angles α and β in FIG. 2B) with respect to the right in the horizontal direction is quantized, and the quantized value sets are changed to different maneuvers. Give each one a different label. As for the end points, for the two wrinkle lines that break, the angle (angle α in FIG. 2C) in the horizontal direction in the direction in which the line extends from the break points is quantized, and the quantized values become different maneuvers. Give each one a different label. For the branch point, the angle of the branched line with respect to the horizontal right direction (angle α in FIG. 2D) and the angle with respect to the horizontal direction in the direction in which the line branched from the branch point extends (angle β in FIG. 2D). ), And assigns different labels to minutiae with different sets of quantized values.

  By performing labeling in this way, the minutiae can be expressed by labeled coordinates (c, x, y). Here, c is one of the labels, and x and y are the x and y coordinates of the minutiae, respectively.

  FIG. 3 is a diagram for explaining how to arrange the minutiae expressed as labeled coordinates in the first embodiment (A), and FIG. 3B for explaining that the same character string is generated from the arrangement of different minutiae. is there.

  The character string generation unit 25 makes a character string by arranging a set of minutiae that is a set of labeled coordinates according to a specific rule. Specifically, a certain point is set as a center point, and feature points are arranged as a character string clockwise from the center point. The diagram on the left side of FIG. 3 (A) represents the arrangement of the minutiae with labeled coordinates. O indicates a center point. Then, as shown in the diagram on the right side of FIG. 3A, the character string in this case is coded as ebadcebaea.

  By using character strings in this way, it is possible to apply collation with one-dimensional character string information, as well as approximate character string collation, instead of collation with two-dimensional information using an image as it is, Collation by a method with a small amount of calculation can be realized.

  Note that FIG. 3B shows a set of labeled coordinates that have the same character string (code: ebadcebaea) despite the arrangement of minutiae different from that in FIG. In this way, since there are an infinite number of labeled coordinate sets that can be the same character string, the original palm information cannot be restored from this character string. Therefore, this conversion can be performed from a set of labeled coordinates to a character string, but cannot be converted from a character string to a set of labeled coordinates, and is a one-way conversion.

The generated string, as the feature data, is transmitted from the feature data transmitter 15 of the mobile communication terminal 3 _l to the feature data receiving unit 31 of the server 5.

  The character string changes due to noise during extraction of minutiae, extraction omission, erroneous extraction, and the like. Therefore, collation cannot be performed by simple character string comparison. In this embodiment, the approximate matching unit 35 of the server 5 performs the approximate matching by deriving the edit distance for the character string proposed by Wagner et al., Thereby replacing the approximate matching of the set of minutiae.

  First, the editing distance of Wagner et al. Will be described.

  Let A be a finite character string consisting of characters, and Λ be an empty string. The character string length of A is represented by | A |.

  A character string pair (a, b) having a length of 1 or less and both of which is not simultaneously Λ is called an editing operation. (a, b) is also expressed as a → b.

  Further, the fact that the character string B is a character string resulting from applying the editing operation a → b to A means that A = σaτ and B = σbτ hold for a certain character string σ, τ. At this time, it is expressed as A⇒B via a → b. Regarding a → b, it is called a change operation when a ≠ Λ and b ≠ Λ, a delete operation when b = Λ, and an insert operation when a = Λ.

In addition, the column s _1, s ₂ of the editing operation of the S, ···, and s _m. A = A is _0, B = B _m, and, 1 ≦ i ≦ m for A _i-1 ⇒A _i via s column A ₁ of _i become such string, A _2, ···, A _m Is written as A⇒B via S.

  Further, γ is an arbitrary cost function that gives a positive real number γ (a → b) for each editing operation a → b. Γ is a distance function that satisfies the following expressions (1), (2), and (3). For the column S of editing operations, the following expression (4) is expressed.

  Furthermore, δ (A, B) is defined as the following equation (5), which is called the edit distance between A and B.

  Furthermore, A <i> and B <i> represent the i-th characters of A and B, respectively. Also, let A (i) represent the character string from 1 to i of A, B (j) represent the character string from 1 to j of B, and D (i, j) Then, the following equation (7) holds (Wagner, RA, 1 other author, “The String-to-String Correction Problem”, J. ACM, Vol. 21, Issue 9, pp.168-173 ( (See Theorem 7 in 1973)).

  FIG. 4 is a diagram illustrating an algorithm for obtaining the edit distance in the first embodiment.

  The edit distance can be obtained by the algorithm shown in FIG. D [i, j] is an edit distance to be obtained.

The approximate collation unit 35 collates the feature data with the data stored in the database 33 using such approximate character string collation. The verification result determining section 37 makes a determination based on the comparison result, the determination result transmitting unit 39 transmits the determination result to the mobile communication terminal 3 _l.

  When the determination result receiving unit 17 of the mobile communication terminal receives the determination result, the determination result display control unit 19 performs display control processing on the display unit 21 based on the determination result.

  The feature data transmission unit 15 may transmit the character string together with the identification information to the server, and the approximate matching unit 35 performs matching with the template corresponding to the identification information among the data stored in the database 33. You may do it.

  In this embodiment, the verification system 1 in FIG. 1 is a palm fortune telling system.

The palm fortune-telling system can be realized by performing the same processing as in the first embodiment. Then, the server 5 determines which type of palm fortune telling the palmprint image of the user of the mobile communication terminal 3 _l is based on the collation result, and transmits the determination result to the mobile communication terminal 3 _l .

For Palmistry system, the database 33 of FIG. 1, the data that is not based biometric information of the user of the mobile communication terminal 3 _l in restorable information is stored. Therefore, in the case of Palmistry system of this embodiment, the mobile communication terminal 3 _l of a user does not need to notify the identity of the recoverable information and user their biometric information to the server 5, the server 5 You can use the service with confidence without worrying about information leakage.

This embodiment, the feature point extraction unit 23 of the mobile communication terminal 3 _l image processing unit 13 in FIG. 1, and performs the feature point extracted using Haar-like features.

  With the Haar-like feature, more local features that are similar to the features obtained by the Haar wavelet transform can be calculated. Recently, object recognition technology using Haar-like features (especially facial image recognition) has been actively studied (for example, P. Viola, 1 other author, “Rapid Object Detection using a Boosted Cascade of Simple Features”, IEEE (See CVPR, 2001). Many of the existing methods for calculating image features (vectors) based on Haar-like features determine the Haar-like filter to be applied by machine learning so that the detection accuracy of training data is high. However, in order to apply to a wide variety of object recognition such as a pictorial book database, it is necessary to construct Haar-like features for each object, and there is a problem that the search time becomes enormous.

  The inventors have been engaged in research on a new wavelet filter learning method suitable for face image detection and recognition using free parameters of a lifting wavelet. The lifting wavelet transform is used for improving the compression ratio of signals and images and as a high-speed calculation technique.

  In this embodiment, the feature point extraction unit 23 uses a lifting Haar-like feature capable of calculating many features at high speed. This is a feature extraction method that can extract a new feature (useful for object recognition) from a feature image obtained by an existing Haar-like feature by using a weight parameter.

  The Haar-like feature can calculate various local features of the image by changing the filter configuration and resolution. This means that the filter for extracting the optimum feature differs depending on the resolution of the image, the type of object in the image, and the like. Therefore, in order to efficiently extract image features, it is necessary to construct (or find) a filter suitable for the target database image.

  FIG. 5 is a diagram illustrating a typical example of a Haar-like filter in the third embodiment. FIG. 5A is an edge feature, and FIG. 5B is a line feature. 5 (C) shows center-surround features, and FIG. 5 (D) shows special diagonal line features.

  In the lifting Haar-like feature in this embodiment, first, a feature image converted by an existing Haar-like feature is prepared. A new feature image is constructed by combining these feature images. At this time, the weight parameter used for the combination is called a lifting parameter. Here, an existing Haar-like feature and a feature image obtained thereby are assumed to be an initial filter and an initial feature image, respectively. A new feature image obtained using the lifting parameters is called a lifting feature image.

  FIG. 6 is a diagram illustrating an outline of a lifting parameter learning method according to the third embodiment. In order to construct a lifting Haar-like feature suitable for object recognition, a lifting parameter is calculated by learning. Here, learning means obtaining a lifting parameter for calculating a feature that minimizes intra-class variance and maximizes inter-class variance. Specifically, for example, if there is a database of images of certain flowers, and they are classified by type (class), the features within those classes will be separated and the features within the classes will approach Thus, lifting parameters are determined.

  Next, the accuracy evaluation of the verification system 1 in FIG. 1 will be described. Through experiments, we examine the accuracy of the identity verification method using palm images. First, the accuracy of identity verification is formulated. Then, under the assumption that the minutiae is appropriately extracted from the palm image, the accuracy of the method based on the approximate character string matching is evaluated.

In this experiment, the mobile communication terminal 3 _l of users, in advance and registering his biometric information in the database 33 of the server 5 as a template. As a procedure for identity verification, first, the feature data transmission unit 15 transmits identification information and feature data for identifying a template to the server 5, and then the server 5 is a template identified from the identification information. And the received feature data are checked to determine whether the person corresponds to the identification information.

  FAR (False Acceptance Rate) and FRR (False Rejection Rate) are used as indicators for the accuracy of identity verification. The FAR is a probability that a user is determined to be a person corresponding to the identification information in spite of submitting identification information that does not correspond to himself / herself in a certain identity verification attempt. The FRR is a probability that a user is determined not to be a person corresponding to the identification information even though the identification information corresponding to the user has been submitted in a certain identity verification attempt.

FAR and FRR are formulated as the following formulas (8) and (9). Here, U = {u _i | i∈I} corresponds to a set of persons who are candidates for identity verification, and i and S _i = {s _i ^j | j∈J} correspond to u _i , respectively. Identification information and feature data. Further, S = {S _i | i∈I}. u _i shall submit s _i ^j selected with equal probability from S _i in a trial with identity verification. Here, f (i, s) is defined as 1 when the server 5 determines that the user is a person corresponding to i by submitting s _i ^j , and is defined as 0 in other cases.

  We conducted an experiment to detect maneuvers from palm images actually taken with a mobile phone.

  FIG. 7 is a diagram showing a flow of a series of processes for detecting a minutia from a palm image (original image (A)) actually taken with a mobile phone in the fourth embodiment, and the feature point extraction unit 23 of FIG. Corresponds to the action. FIG. 7A shows an original image. The original image is color, and first, grayscale conversion is performed. FIG. 7B shows a grayscale image. Next, the edges are extracted using the Canny method (see J. F. Canny, “A computational approach to edge detection”, IEEE Transaction on Pattern Analysis and Machine Intelligence, pp. 679-698). FIG. 7C is a diagram showing the extracted edges. Then, minutiae candidate points are detected by template matching. FIG. 7D is a diagram illustrating detected minutiae candidate points. As can be confirmed in FIG. 7D, many points called pseudo-maneuvers that are not correct maneuvers are detected. This is caused by using an input image containing a lot of noise. In this experiment, in order to investigate the influence of the processing after extracting a set of minutiae on the accuracy of identity verification, the experiment is performed using a minutiae visually detected from a palm image.

  As samples used in this experiment, 40 palm images of 20 different persons were collected in total, 2 each. The image resolution is 4000 × 3000 pxi. In this experiment, the number of types of minutiae was 33, and the number of minutiae extracted for one image was about 20.

  When the extracted minutiae set is used as a feature point set, character string conversion is performed, and a character string for one of the palm images of each person is used as a template, and the edit distance between the submitted character string and the template is less than the threshold The person corresponding to the template is determined. However, for normalization, the value of the edit distance is divided by the length of the longer character string, and the threshold value is expressed as a ratio to the length of the longer character string. By changing the threshold value, FAR and FRR were obtained by the equations (8) and (9).

  FIG. 8 is a diagram illustrating experimental results of FAR and FRR for the accuracy evaluation of the matching system 1 in the fourth embodiment. The horizontal axis represents the threshold value, and the vertical axis represents FAR and FRR. From this, FAR and FRR are in a trade-off relationship, and when the threshold value is 0.5, FAR = 0.042 and FRR = 0.05, and the accuracy when FAR and FRR are minimum is clarified. Further, since the FAR value is important for attacks such as impersonation, when the FAR is important, when the threshold value is 0.4, FAR = 0.000 and FRR = 0.15, and this time is ideal accuracy. From this, it became clear that biometric authentication using the present invention is practical when feature points are extracted appropriately.

It is a block diagram which shows the structure of the collation system 1 which is one of embodiment of this invention, and the flow of a process. In Example 1, it is the figure (A) explaining the intersection, the end point, and the branching point extracted as a minutia, and the figure (B, C, D) which shows the declination in each point. In Example 1, it is the figure (A) explaining how to arrange the minutiae represented by the coordinate with a label, and the figure (B) explaining that the same character string is produced | generated from arrangement | positioning of a different minutia. In Example 1, it is a figure which shows the algorithm which calculates | requires edit distance. In Example 3, it is a figure which shows the typical example of a Haar-like filter, FIG. 5 (A) is an edge feature (Edge features), FIG.5 (B) is a line feature (Line features), FIG. 5C shows center-surround features, and FIG. 5D shows special diagonal line features. In Example 3, it is a figure which shows the outline of the learning method of a lifting parameter. In Example 4, it is a figure which shows a series of processes which detect a minutia from the palm image (original image (A)) actually image | photographed with the mobile phone, FIG. 7 (A) is a figure which shows an original image, 7 (B) is a diagram showing a grayscale image, FIG. 7 (C) is a diagram showing extracted edges, and FIG. 7 (D) is a diagram showing detected minutiae candidate points. In Example 4, it is a figure which shows the experimental result of FAR and FRR about evaluation of the precision of the collation system 1. FIG.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Collation system, 3 ₁ ,..., 3 _n mobile communication terminal, 5 server, 11 image acquisition unit, 13 image processing unit, 15 feature data transmission unit, 17 determination result reception unit, 19 determination result display control unit, 31 Feature data reception unit, 33 database, 35 approximate collation unit, 37 collation result determination unit, 39 determination result transmission unit

Claims (7)

In a collation system comprising a portable communication terminal having an image photographing function and a server for performing a collation process based on image information obtained by photographing with the image photographing function of the portable communication terminal,
The mobile communication terminal is
Image acquisition means for acquiring biological information as image data;
Image data obtained by shooting is converted into Haar-like features into a plurality of feature images, combined with weighted lifting parameters to generate new feature images , and feature data based on the new feature images Image processing means for generating the image in one direction;
Feature data transmitting means for transmitting the feature data to the server,
The server
Feature data receiving means for receiving the feature data;
A collation system comprising approximate collation means for performing an approximate collation process on the feature data and data stored in a database. The image data is palm print image data,
The image processing means includes
Feature point extraction means for extracting feature points from the image represented by the palmprint image data;
A character string generation means for generating a character string from the extracted feature points;
The feature data transmitting means transmits the character string to the server as the feature data.
The approximate matching means performs an approximate character string matching process.
The collation system according to claim 1.   The collation system according to claim 2, wherein the character string generation unit arranges feature points as character strings in a clockwise or counterclockwise direction from the center point with one point of the image represented by the palm print image data as the center point. The data stored in the database is not based on information capable of restoring the biological information of the user of the mobile communication terminal, and relates to a fortune telling service,
The server
Verification result determination means for determining the verification result of the approximate verification means;
A determination result transmitting means for transmitting the determination result by the collation result determining means to the mobile communication terminal,
The mobile communication terminal is
Determination result receiving means for receiving the determination result;
The collation system according to claim 2 or 3 , further comprising determination result display control means for displaying the determination result on a display means. A collation method in a collation system comprising a mobile communication terminal having an image photographing function and a server that performs collation processing based on image information obtained by photographing with the image photographing function of the mobile communication terminal,
The image processing means of the portable communication terminal converts the image data obtained by photographing biometric information into Haar-like features to form a plurality of feature images, weighted with lifting parameters, and combined to create a new feature image Generating and generating feature data in one direction based on the new feature image ;
The feature data transmitting means of the mobile communication terminal transmits the feature data to the server;
The feature data receiving means of the server receives the feature data;
A matching method, comprising: an approximate matching unit of the server performing an approximate matching process on the feature data and data stored in a database. The program for functioning a computer as an image processing means in any one of Claim 1 to 4 . A computer-readable recording medium on which the program according to claim 6 is recorded.