JP2006133945A - Collation device, collation method, and collation program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a collation device which does not require a user to consciously make specific actions, further prevents impersonation, and executes pre-collation processing with less resources. <P>SOLUTION: The collation device selects a guiding method, which makes a user's specific part move, from a plurality of items by means of a guiding movement selection part 110, makes the specific part move by the guidance of a movement guiding part 120 by using the selected guiding method, photographs the user moving the specific part with a camera 210, estimates the movement area of the specific part from the time-series photographed images by means of a movement part estimation part 330, extracts a collation object area based on the specific part estimated from the position of the estimated movement area by means of a collation object area extraction part 350, and collates the extracted area by using the collation object area by means of a collation part 360. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a collation device, a collation method, and a collation program that prevent impersonation.

As a conventional collation device, for example, a device that collates information on an image of a predetermined part of a living body such as a face registered in advance with information on an image of the same part imaged by an imaging unit such as a camera. There is a device that prescribes a specific operation of closing and opening a predetermined time, and collates face images on the condition that the specific operation is detected (see, for example, Patent Document 1).
JP 2002-251614 A

  However, in such a conventional collation device, if the user cannot perform a specific operation defined in order to prevent impersonation by a photograph or the like, the user must repeat the operation many times. There was a problem that the load increased.

  In addition, since processing such as detection of a specific part, contour detection of a specific part, operation type determination, etc. is required before starting the verification process, the amount of processing increases. Then, there was a problem that it was difficult to realize in real time.

  The present invention has been made to solve such a problem, and it is not necessary for the user to be aware of a specific operation, and impersonation can be more reliably suppressed, and before starting collation. It is an object of the present invention to provide a collation apparatus, a collation method and a collation program that can collate with a small amount of resources.

  The collation device of the present invention includes a guidance operation selection unit that selects a guidance method for causing movement at a specific part to be authenticated from a plurality of items prepared in advance, and a guidance method selected by the guidance operation selection unit. An action guiding unit that performs guidance so as to cause movement in a specific part of the authentication target, an imaging unit that images the authentication target, and the authentication target that is guided by the motion guiding part to move the specific part. A motion part estimation unit that estimates a motion region of the specific part from a time-series image captured by the imaging unit, and an operation content determination that determines whether or not the motion of the specific part is an operation content guided by the motion guide unit And the position of the specific part of the authentication target from the position of the motion region estimated by the motion part estimation unit, and the verification target of the authentication target based on the estimated specific part A collation target region extracting unit for extracting a region, a collation unit for collating the collation target region extracted by the collation target region extracting unit with an image captured in advance, and a collation result output unit for outputting a collation result of the collation unit With.

  According to this configuration, it is generated by performing an interactive operation between the device and the authentication target, and selecting a guidance method for causing movement at a specific part of the authentication target from a plurality of items prepared in advance. Since the verification region to be authenticated is determined based on the movement of the specific part to be authenticated, impersonation with a photograph or the like can be more reliably prevented. In addition, since the part specification can be performed at the same time when the user's motion is detected, the preprocessing for collation can be performed with a small amount of processing.

The collation apparatus of the present invention is the collation device according to the above-described configuration, wherein the moving part detection unit calculates a horizontal differential value in a pixel in the time-series image, and a vertical differential in the pixel in the time-series image. A vertical differential calculation unit that calculates a value, a time differential calculation unit that calculates a differential value in the time direction of the pixel of the time-series image, and each horizontal, vertical, and temporal differential value of the pixel of the time-series image. A vector calculation unit that calculates a motion vector based on the vector.

  According to this configuration, since the movement of the specific part to be authenticated can be detected at the same time as the part specification using the optical flow calculation algorithm with a small amount of processing, it is possible to perform preprocessing of matching with a small amount of processing. it can.

  The collation device of the present invention is characterized in that, in the configuration described above, the motion guide unit includes a character display unit that displays a character, and a character motion unit that outputs the character that performs a motion that causes a specific part to be authenticated. Prepare.

  According to this configuration, since the verification area is detected while talking with the character displayed on the screen, it is possible to perform verification that prevents spoofing without being aware of the authentication target.

  The collation apparatus of this invention is equipped with the light emission part which light-emits a flash in the said structure.

  According to this configuration, it is possible to set the face area for verification by detecting the eye movement of the authentication target that occurs when the flash is emitted, and it is possible to perform verification that prevents spoofing without being conscious of the authentication target.

  In the verification apparatus according to the present invention, in the configuration, the verification target region is a face, and the specific part is a part of the face.

  According to this configuration, it is possible to prevent impersonation without being conscious of the authentication target by performing face matching.

  The collation method of the present invention includes a step of selecting a guidance method for causing movement in a specific part to be authenticated from among a plurality of items prepared in advance, and a movement to a specific part to be authenticated by the selected guidance method. A step of performing guidance so as to generate, a step of imaging the authentication target that is guided and moving the specific part, and an operation region of the specific part is estimated from a time-series image captured by the imaging unit A step of determining whether or not the motion of the specific part is the induced motion content, estimating the position of the specific part to be authenticated from the estimated position of the motion region, and estimating the specific part A step of extracting the verification target area based on the authentication, a step of collating the extracted verification target area with a pre-captured image, and outputting a verification result And a step.

  According to this configuration, it is generated by performing an interactive operation between the device and the authentication target, and selecting a guidance method for causing movement at a specific part of the authentication target from a plurality of items prepared in advance. Since the verification region to be authenticated is determined based on the movement of the specific part to be authenticated, impersonation with a photograph or the like can be more reliably prevented. In addition, since the part specification can be performed at the same time when the user's motion is detected, the preprocessing for collation can be performed with a small amount of processing.

The collation program of the present invention includes a function for selecting, from a plurality of items prepared in advance, a guidance method for causing a computer to cause a movement in a specific part to be authenticated, and a specific part to be authenticated by the selected guidance method. A function of performing guidance so as to cause movement, a function of imaging the authentication target that is guided and moving the specific part, and a time series image captured by the imaging unit of the specific part. A function for estimating an action area, a function for estimating the position of the specific part to be authenticated from the estimated position of the action area, and a function for determining whether the action of the specific part is an action content induced. A function of extracting the verification target area of the authentication target based on the estimated specific part, a function of matching the extracted verification target area with an image captured in advance, and a verification result To a function and you want to run.

  According to this configuration, it is generated by performing an interactive operation between the device and the authentication target, and selecting a guidance method for causing movement at a specific part of the authentication target from a plurality of items prepared in advance. Since the verification region to be authenticated is determined based on the movement of the specific part to be authenticated, impersonation with a photograph or the like can be more reliably prevented. In addition, since the part specification can be performed at the same time when the user's motion is detected, the preprocessing for collation can be performed with a small amount of processing.

  As described above, according to the collation device, the collation method, and the collation program of the present invention, a guidance method for performing interactive processing between the apparatus and the authentication target and causing a motion to occur in a specific part of the authentication target is prepared in advance. By selecting from a plurality of selected items and guiding the selected part to cause movement in the specific part of the authentication target, the authentication target is naturally generated in the process of operating the device. Since the movement can be detected and collated, the load on the authentication target can be reduced and impersonation can be more reliably prevented. In addition, since the part can be specified when detecting the movement of the authentication target, collation can be performed with a small amount of pre-processing, and therefore impersonation using a low-resource device such as a mobile phone can be prevented. Can do.

  Embodiments of the present invention will be described below with reference to the drawings.

  FIG. 1 is a block diagram of a collation apparatus according to an embodiment of the present invention.

  In FIG. 1, this collation apparatus has the user's action to be authenticated that is stored and prepared in advance in the user action database 140 such as “flash emission”, “speak to character”, “make the character imitate”. A guidance operation selection unit 110 that randomly selects one or a plurality of items from among a plurality of items that induce guidance, and a guidance method selected by the guidance operation selection unit 110, for example, flashing light or an authentication target By displaying a character to be talked to the user, the action guiding unit 120 that guides the user to move the specific part of the user, for example, a part of the face, that is, the eyes, and the collation result are displayed on the display. A display unit 100 including a collation result output unit 130 and a camera 210 which is an image capturing unit that captures an image of the user's face. Force means 200, an image input unit 310 that inputs an image of the user's face, a motion part detection unit 320 that detects a motion vector from a time-series image captured by the camera 210, and a motion in the face from the motion vector A motion part estimation unit 330 that estimates the region of the part, a motion content determination unit 340 that determines whether or not the user's motion is motion content derived from the motion vector of the region detected by the motion part detection unit 320, and a motion part estimation A collation target region extraction unit 350 that estimates a user's specific part, that is, an eye from the position of the motion region estimated by the unit 330, and extracts the user's collation target area, such as a face, based on the estimated specific part; A collation unit 360 is provided that collates the face area image extracted by the area extraction unit 350 with a face image captured in advance to determine whether the person is the person himself or herself. And a certification unit 300.

  Next, the operation of the collation device according to the present embodiment will be described.

  FIG. 2 is a flowchart showing a processing procedure of the collation device according to the present embodiment. FIG. 3 is an explanatory diagram showing an example of a user's face image captured when the flash is emitted by the collation device according to the present embodiment. FIG. 4 is a diagram illustrating an example of pixels used when calculating a motion vector of one pixel of the collation device according to the present embodiment. FIG. 5 is a diagram illustrating an example of motion vectors calculated by the collation device according to the present embodiment. FIG. 6 is a diagram illustrating an example of a face area extracted by the collation device according to the present embodiment.

  First, the guidance operation selection unit 110 randomly selects one of a plurality of items in the guidance operation database 140 that is prepared by holding a guidance method for causing the user to perform an action. The selection items include, for example, “flash emission”, “conversation with character”, “simulation of character movement”, and the like. When the guidance operation selection unit 110 selects, for example, “flash emission” (step S200), the operation guidance unit 120 causes the flash to emit light (step S201), and the user is induced to blink. At this time, since the user moves his eyes as shown in FIGS. 3A to 3D, the operation guiding unit 120 notifies the start of processing at the same time as the flash is emitted. Here, FIGS. 3A and 3C show a state where the eyes are open, and FIGS. 3B and 3D show a state where the eyes are closed. When the flash is emitted, the movement of the eye area increases because the user blinks. Since the camera 210 captures an image of the user, when receiving a processing start notification from the motion guiding unit 120, the image input unit 310 inputs an image as shown in FIG. 3 frame by frame (step S202). The motion vector detection unit 320 calculates an optical flow in the image for the time-series image in which the user is blinking (step S203).

  When the optical flow is calculated, a motion vector 501 as indicated by an arrow in FIG. 5 is obtained for each pixel with induced motion, that is, each pixel of the eye contour. Therefore, the motion region estimation unit 330 estimates and extracts (cuts out) an area where the optical flow exists as an eye area (step S204). Since the motion of the upper contour of the eye when blinking is a vertical motion, if the optical flow of the motion region extracted by the motion region estimation unit 330 is a vertical optical flow (step S205, Yes), the motion The content determination unit 340 outputs the position of the eye region to the verification target region extraction unit 350. On the other hand, if the optical flow of the motion region extracted by the motion part estimation unit 330 is not the upper or lower optical flow (No in step S205), the process ends without matching. The target area extraction unit 350 extracts a face area based on the position of both eye areas input from the operation content determination unit 340 (step S206). For example, as shown in FIG. 6, the length twice as wide as the eye width 601 is set to the horizontal width and the vertical width of the face area 605, and a point vertically below the eye width half 603 from the center 602 of both eyes is the face area center 604. The rectangular area thus extracted is extracted as a face area 605. Finally, the collation unit 360 compares the face image extracted in advance with the face region extracted by the target region extraction unit 350, thereby collating whether or not the person is the person (step S207). At this time, the matching algorithm is not particularly limited, and for example, matching using a registered face template may be used.

  Here, the guidance method by the motion guidance unit 120 may be anything as long as the motion can be derived from the user. In other words, the flash may be emitted to cause the user to blink, to have a conversation with the character, or to imitate the action of the character and allow the user to move the face. When talking with the character, the part where the movement occurs is the area of the user's mouth. Therefore, in the face area extraction performed by the face area extracting unit 350, the width of the mouth is twice the width of the mouth and the width of the mouth. For example, a rectangular area centered on a point that is perpendicular to the center and half the interval between the mouths may be extracted as a face area. When the face can be moved by imitating the movement of the character, the part where the movement occurs is the entire head of the user. Therefore, the detection of the face area performed by the target area extraction unit 350 is centered on the center of the entire head. Alternatively, a rectangular area in which the distance from the center to the upper end is set to the horizontal width and the vertical width may be extracted as the face area.

  In addition, the guidance operation selection unit 110 may select a plurality of processing items. For example, two of “flash emission” and “speak to character” may be selected. At this time, the operation guiding unit 120 performs two processes in order.

  The means for estimating the specific part that moves is not limited to the optical flow. For example, a difference between two temporally continuous images may be taken and a region with a large difference may be estimated as a moving part.

  As an equation for calculating the optical flow, for example, an equation using a light gradient method of processing is used. The method of optical flow calculation by the gradient method is shown below. Points in the image

の、時刻における輝度を

Of brightness at time

とし、この点が時刻

And this point is the time

に点

Point to

に移動すると仮定すると、

Assuming that

が得られる。

Is obtained.

  Luminance

が

But

に関して滑らかに変化したとすると、（数５）の左辺をテイラー展開でき、

If the left side of (Equation 5) is Taylor expanded,

を得る。

Get.

  here,

は微小であるので省略できる。（数８）を

Is small and can be omitted. (Equation 8)

で割り、

Divide by

とした極限をとり整理して、

Organize the limit as

を得る。

Get.

  differential

は画像から計算される。
（数１）の近傍領域を

Is calculated from the image.
The neighborhood area of (Equation 1)

とすると、近傍領域（数１４）で動きが滑らかに変化するならば、（数１２）で表される拘束式に関する

Then, if the motion changes smoothly in the neighborhood region (Equation 14), the constraint equation represented by (Equation 12)

の誤差が最小になる。

Error is minimized.

が最小になるとき、

When is minimized

であるから、

Because

を得る。

Get.

  Organize (Equation 18)

が得られる。

Is obtained.

  (Equation 19)

について解き、

Solve about

を得る。

Get.

  In order to calculate the optical flow from (Equation 21), the spatial differentiation of luminance

と、時間微分

And time derivative

を計算する必要がある。これは、画像の格子点上の１階差分を用いて求める。１階差分の計算は、図４に示すような４個の画素が交差する場所において２枚の連続する画像の中間点で行う。

Need to be calculated. This is obtained using the first-order difference on the lattice points of the image. The calculation of the first-order difference is performed at an intermediate point between two consecutive images at a place where four pixels intersect as shown in FIG.

をそれぞれ（数７）に対応する添字とすると、（数１３）は

Is a subscript corresponding to (Equation 7), respectively, (Equation 13) becomes

のようになる。

become that way.

  That is, the optical flow calculation can be obtained by using an average of pixel value differences.

  Here, the optical flow calculation method is not limited to the above example.

  As described above, according to the present embodiment, when the motion guiding unit 110 guides the user to cause a motion in a specific part of the user's body, motion vector detection is performed from the time-series image input from the image input unit 310. The unit 320 calculates a motion vector, the collation target region extraction unit 350 estimates and extracts the position of the collation target region from the image based on the motion vector, and the collation unit 360 has captured the extracted collation target region in advance. Since the image is collated, interactive processing can be performed between the apparatus and the user, and impersonation can be prevented by reducing the load on the user during authentication. In addition, a guidance operation selection unit 110 that selects a guidance method for causing a motion at a specific part of the user from a plurality of items prepared in advance, and a motion at the specific part of the user by the selected guidance method. Since the operation guiding unit 120 that performs the guidance is provided, impersonation can be prevented more reliably. Furthermore, when detecting a user's movement, the optical flow calculation algorithm with a small amount of processing can be used to identify a part at the same time as the movement detection, so that collation can be performed with a small amount of preprocessing.

  In this embodiment, the case where the user moves the eyes, estimates the position of the eyes, and extracts the face area used for matching has been described as an example. The estimation is merely an example, and it is of course possible even when other parts of the face, that is, parts such as the mouth move, or when collating with body features other than the face.

  In the above embodiment, the hardware configuration has been described as an example. However, the present invention is not limited to this, and the above functions are configured to be realized by software by the CPU executing a program. May be. That is, the present invention provides a computer with a function of selecting a guidance method for causing a motion in a specific part to be authenticated from a plurality of items prepared in advance, and a specific part to be authenticated by the selected guidance method. A function for performing guidance so as to cause movement, a function for the imaging unit to image an authentication target that is guided and moving the specific site, and a function for estimating an operation region of the specific site from a time-series image captured by the imaging unit A function for estimating the position of a specific part to be authenticated from the position of the estimated motion region, a function for determining whether or not the action of the specific part is a guided action, and an authentication target based on the estimated specific part A collation program for executing a function for extracting a collation target area, a function for collating the extracted collation target area with an image captured in advance, and a function for displaying a collation result. Also, it is possible to achieve the object of the present invention.

  A collation device, a collation method, and a collation program according to the present invention are a plurality of items prepared in advance as a guidance method for performing interactive processing between an apparatus and an authentication target and generating a motion in a specific part of the authentication target. By detecting the natural movement of the authentication target that occurs in the process of operating the device at the time of authentication, the selected guidance method is used to induce the movement to occur in the specific part of the authentication target. Since it is possible to reduce the load on the authentication target, spoofing can be prevented more reliably, and when the movement of the authentication target can be detected, the part can be identified, so there is less verification A personal authentication device that has the effect of being able to be performed by pre-processing and performs authentication for a part of the body, especially a personal authentication device for a low-resource terminal such as a mobile phone. It is useful.

The block diagram which shows the collation apparatus which concerns on embodiment of this invention The flowchart which shows the process sequence of the collation apparatus which concerns on embodiment of this invention. The figure which shows the example of the image of the authentication target imaged with the collation apparatus which concerns on embodiment of this invention The figure explaining calculation of the motion vector of 1 pixel by the motion region estimation part which concerns on embodiment of this invention The figure which shows the example of the motion vector by the motion part estimation part of the collation apparatus which concerns on embodiment of this invention The figure which shows the collation area | region which the collation target area extraction part of the collation apparatus which concerns on embodiment of this invention extracts

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 Display means 110 Guidance motion selection part 120 Motion guidance part 130 Collation result output part 140 Guidance action database 200 Input means 210 Camera 300 Authentication means 310 Image input part 320 Motion vector detection part 330 Motion part estimation part 340 Action content determination part 350 Collation Target area extraction unit 360 verification unit

Claims (7)

A guidance operation selection unit that selects a guidance method for causing movement in a specific part to be authenticated from a plurality of items prepared in advance;
An operation induction unit that performs induction so as to cause movement in a specific part to be authenticated by the induction method selected by the induction operation selection unit;
An imaging unit that images the authentication target;
A motion part estimation unit that estimates a motion region of the specific part from a time-series image in which the imaging unit images the authentication target that is guided by the motion guide part and moves the specific part;
An operation content determination unit for determining whether the operation of the specific part is the operation content induced by the operation induction unit;
A verification target region extraction unit that estimates the position of the specific part of the authentication target from the position of the motion region estimated by the motion part estimation unit, and extracts the verification target region of the authentication target based on the estimated specific part When,
A collation unit that collates the collation target region extracted by the collation target region extraction unit with an image captured in advance;
A collation result output unit for outputting a collation result of the collation unit;
A collation device comprising: The movement part detection unit is
A horizontal differential operation unit for calculating a horizontal differential value in the pixels of the time-series image;
A vertical differential operation unit for calculating a differential value in the vertical direction in the pixels of the time-series image;
A time differential calculation unit for calculating a differential value in the time direction in the pixels of the time-series image;
A vector calculation unit that calculates a motion vector based on the differential values in the horizontal, vertical, and time directions of the pixels of the time-series image;
The collation apparatus of Claim 1 provided with. The motion induction unit is
A character display for displaying the character;
A character action unit for outputting the character for performing a guidance for causing movement in the specific part to be authenticated;
The collation apparatus of Claim 1 or 2 provided with. The motion induction unit is
4. The collation device according to claim 1, further comprising a light emitting unit that emits a flash. The verification target area is a face;
The collation apparatus according to claim 1, wherein the specific part is a part of the face. Selecting from among a plurality of items prepared in advance a guidance method for causing movement in a specific part to be authenticated;
Performing guidance so as to cause movement in a specific part to be authenticated by the selected guidance method;
An imaging unit that images the authentication target that is guided and moves the specific part; and
Estimating a motion region of the specific part from a time-series image captured by the imaging unit;
Determining whether the action of the specific part is the induced action content;
Estimating the position of the specific part of the authentication target from the estimated position of the motion area, and extracting the verification target area of the authentication target based on the estimated specific part;
Collating the extracted region to be collated with an image captured in advance;
Outputting a verification result; and
A collation method comprising: On the computer,
A function of selecting a guidance method for causing movement in a specific part to be authenticated from a plurality of items prepared in advance;
A function to guide the selected part to cause movement in a specific part of the authentication target,
A function for the imaging unit to image the authentication target that is guided and moves the specific part;
A function of estimating an operation region of the specific part from a time-series image captured by the imaging unit;
A function of estimating the position of the specific part to be authenticated from the position of the estimated motion region;
A function of determining whether the action of the specific part is the induced action content;
A function of extracting a verification target region of the authentication target based on the estimated specific part;
A function of collating the extracted region to be collated with an image captured in advance;
A function to display the verification results;
A collation program that executes

Images