JP2001222716A - Personal identification method and personal identification system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a personal identification method and a personal identification system which are capable of simply and accurately identifying a person. SOLUTION: The three-dimensional data of a person face are preliminarily stored as dictionary data in a data storage means, then the position attitude of the three-dimensional data inputted to a three-dimensional data inputting means 1 is detected at the time of identification, a two-dimensional picture for identification and a two-dimensional picture to be identified of the same position attitude as each other are respectively generated from the dictionary data and the three-dimensional data on the basis of the detection results, and a person is identified by comparing the two-dimensional picture for identification with the two-dimensional picture to be identified.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a person authentication method for authenticating a person and a person authentication device using the person authentication method.

[0002]

2. Description of the Related Art
In a security system or the like, image data of a part or the whole of a person is stored in advance, and the person is authenticated using the stored image data and the image data input by the data input unit at the time of authentication. Methods are known.

[0003] Among them, a person authentication method using two-dimensional data, that is, comparing two-dimensional data stored in advance with two-dimensional data input by a two-dimensional data input device such as a TV camera at the time of authentication, to identify a person. There have been many studies on personal authentication methods for authentication.

However, there are the following problems in the person authentication method using the two-dimensional data. That is, the face or body of a person is a three-dimensional object, and the two-dimensional data input by the two-dimensional data input device includes the two-dimensional data stored in advance and the direction of the face, the size of the face on the image, or the illumination. There is a problem that the conditions often do not match.

[0005] As a method of solving this problem, various face orientations must be set in advance in order to correct the face orientation mismatch.
It is conceivable to store dimensional data, but in this case, a problem arises in that a large amount of two-dimensional data must be stored. Further, in order to correct the mismatch of the size of the face on the image, it is conceivable to perform correction (normalization) for matching the size of the face of the stored two-dimensional data and the input two-dimensional data on the image. However, this causes a problem that information about the actual face size is lost. Further, it is conceivable to control the lighting conditions in order to correct the mismatch of the lighting conditions, but this causes a problem that the place where the person can be authenticated is limited.

On the other hand, a person authentication method using three-dimensional data directly, that is, a person authentication method in which three-dimensional data previously stored as dictionary data is compared with three-dimensional data input by a three-dimensional data input device. If this is the case, the problem related to the face orientation as described above is solved. The person authentication method using the three-dimensional data includes, for example, a method of aligning dictionary data with input three-dimensional data and using a distance between the two three-dimensional data as a determination parameter, and a method of three-dimensional data.
There are a method of extracting parts such as a nose and a mouth from the dimensional data and comparing the parts, and a method of expressing both three-dimensional data as parameters representing a shape such as curvature and comparing the parameters.

[0007] However, any person authentication method directly using three-dimensional data has a problem that it is susceptible to noise and it is difficult to extract characteristic parameters.

The present invention has been made in view of the above problems, and has as its object to provide a person authentication method and apparatus capable of easily and accurately authenticating a person.

[0009]

In order to achieve the above-mentioned object, a person authentication method according to the present invention stores three-dimensional data of a part or the whole of a person in advance as dictionary data, A person authentication method for authenticating a person using data and three-dimensional data input by three-dimensional data input means at the time of authentication, wherein a position and orientation of the three-dimensional data input by the three-dimensional data input means are detected. Then, based on the detection result, the authentication data of the same position and orientation are obtained from the dictionary data and the input three-dimensional data.
A two-dimensional image and a two-dimensional image for authentication are respectively generated, and the two-dimensional image for authentication and the two-dimensional image for authentication are compared to authenticate a person.

According to this method, since the two-dimensional image for authentication and the two-dimensional image for authentication are both generated at the same position and orientation from the three-dimensional data, the orientation of the face and the face on the image are compared. The size and the lighting conditions can be matched respectively. Also, what is actually used for comparison is
Since both the two-dimensional image for authentication and the two-dimensional image for authentication are two-dimensional images, it is possible to reduce noise that would have occurred when three-dimensional data was directly used, and
There is no need to extract characteristic parameters, and a person can be easily and accurately authenticated.

It is preferable that a shading image or a texture mapping image is used as the two-dimensional image for authentication and the two-dimensional image for authentication. According to this, a two-dimensional image can be easily generated from three-dimensional data by a normal computer graphics technique, and a person can be more easily and accurately authenticated.

On the other hand, a person authentication apparatus according to the present invention comprises a three-dimensional data input means for inputting three-dimensional data on a part or the whole of a person, and three-dimensional data on a part or the whole of a person as dictionary data. Storage means for storing in advance, position and orientation detection means for detecting the position and orientation of the three-dimensional data input by the three-dimensional data input means, and the dictionary data based on the detection result of the position and orientation detection means. A two-dimensional image generating means for respectively generating a two-dimensional image for authentication and a two-dimensional image for authentication having the same position and orientation from the input three-dimensional data;
Image comparing means for comparing the two-dimensional image and the two-dimensional image for authentication with each other to authenticate a person.

According to this apparatus, the above-described person authentication method can be easily and reliably implemented.

[0014]

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

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

In FIG. 1, three-dimensional data input means 1
Is capable of inputting three-dimensional data having the accuracy and resolution required for face recognition of a person. The three-dimensional data input means 1 includes, for example, a stereo method for stereoscopically recognizing a person's face by photographing with two or more cameras installed at different places, a thin slit at an angle of about 45 degrees with respect to the person's face. Project an image and observe the image from the specular reflection direction, or illuminate the face of the person's face with a thin slit-like light beam so as to cut it, and observe the shape of the cutting line generated on the surface from the side 3D recognizes a person's face by observing the difference in the distance between the lattices and the interference fringe pattern caused by mutual rotation. Various types using a moire method or the like can be given. The three-dimensional data input to the three-dimensional data input means 1 is, in addition to the three-dimensional shape data, associated with the shape data if necessary.
It may include dimensional data.

The data storage means 2 stores in advance, as dictionary data, the three-dimensional data of the face of the authenticated person previously input by the three-dimensional data input means 1 or another three-dimensional data input means. . It is preferable that the dictionary data of these authentication persons have the same position and orientation. Therefore, the dictionary data uses a jig for fixing the head of the authentication person, or converts the input three-dimensional data into the position and orientation detection means described later. It is desirable to make corrections so that the position and orientation are constant by 3 or the like. Note that a hard disk or a magneto-optical disk may be used as the data storage means.

The position / orientation detecting means 3 is based on dictionary data previously stored in the data storage means 2 and three-dimensional data (hereinafter referred to as measurement data) inputted by the three-dimensional data input means 1 at the time of authentication. It detects the position and orientation of the measurement data. For example, as a method of estimating and detecting the position and orientation from only three-dimensional data, "" UsingSp
in Images for Efficient Object Recognition in Clut
tered 3D Scenes ", IEEE Trans.on PATTERN ANALYSIS AN
D MACHINE INTELLIGENCE, VOL. 21, NO. 5, 1999 ". As a method of estimating and detecting the position and orientation from a three-dimensional model of contour lines such as eyes, lips, and eyebrows and two-dimensional data, “HeadPose Determinat
ion from One Image Using a Generic Model ", 3rd.Int
l.conf.on Face & Gesture Recognition, 1998 ”.

The two-dimensional image generating means for authentication 4 and the two-dimensional image generating means for authentication 5 generate the two-dimensional image data for authentication of the same position and orientation from the dictionary data and the measurement data based on the detection result of the position and orientation detecting means 3. A two-dimensional image and a two-dimensional image for authentication are generated. Regarding the generation of the position and orientation of the two-dimensional image for authentication and the two-dimensional image for authentication,
Based on the detection result of the position and orientation detection means 3, the observation position and orientation of the generated image may be adjusted to the shooting position and orientation of the dictionary data or measurement data, or may be adjusted to an intermediate position between the two.

Preferably, a shading image or a texture mapping image is used as the two-dimensional image for authentication and the two-dimensional image for authentication. According to this, 3 is obtained by the usual computer graphics technique.
A two-dimensional image can be easily generated from the dimensional data. When a texture mapping image is used as the two-dimensional image for authentication and the two-dimensional image for authentication, each point of the three-dimensional data input by the three-dimensional data input means 1 can be associated with the two-dimensional image. It is necessary to be. For example, an input device such as VIOL700 manufactured by Minolta that can simultaneously input three-dimensional data and two-dimensional data may be used as the three-dimensional data input means.

The image comparing means 6 compares the two-dimensional image for authentication generated by the two-dimensional image generating means for authentication 4 and the two-dimensional image for authentication generated by the two-dimensional image generating means for authentication 5 with each other. The person is authenticated by comparison, and the authentication result is output to a monitor or the like (not shown).
For example, a method described in a document such as “Unlock control system using face image collation”, IEICE, Vol. 80D-II, pp. 2203-2208, 1997 may be used for this authentication.

Next, the operation of the person authentication apparatus shown in FIG. 1 will be described.

First, dictionary data is created in advance.
That is, the three-dimensional data of the face of the authenticated person is previously input by the three-dimensional data input means 1 and stored in the data storage means 2 as dictionary data, and this operation is repeated by the required number of authenticated persons.

Next, at the time of authentication, three-dimensional data input means 1
After inputting the three-dimensional data of the face of the person to be authenticated as the measurement data, the position and orientation detection means 3 reads the dictionary data stored in the data storage means 2 in advance,
The position and orientation of the measurement data are detected from the measurement data.

The two-dimensional image generating means for authentication 4 and the two-dimensional image generating means for authentication 5 authenticate the same position and orientation from the dictionary data and the measurement data based on the detection result of the position and orientation detection means 3. A two-dimensional image for authentication and a two-dimensional image for authentication are generated.

The image comparing means 6 compares the two-dimensional image for authentication generated by the two-dimensional image generating means for authentication 4 with the two-dimensional image for authentication generated by the two-dimensional image generating means for authentication 5. Are compared with each other to authenticate the person, and the authentication result is output to a monitor or the like (not shown).

As described above, the two-dimensional image for authentication and the two-dimensional image for authentication are both generated at the same position and orientation from the dictionary data and the measurement data, which are three-dimensional data. The size of the upper face and the lighting conditions can be matched. Further, since both the two-dimensional image for authentication and the two-dimensional image for authentication used for the actual comparison are two-dimensional data, it is possible to reduce noise that would have occurred when the three-dimensional data was directly used, and There is no need to extract characteristic parameters, and a person can be easily and accurately authenticated.

[0028]

According to the first aspect of the present invention, a person is authenticated by comparing a two-dimensional image for authentication and a two-dimensional image for authentication with the same position and orientation generated from three-dimensional data. The face orientation, the size of the face on the image and the lighting conditions can be matched,
Noise can be reduced, and there is no need to extract characteristic parameters, and a person can be easily and accurately authenticated.

According to the second aspect of the present invention, a two-dimensional image can be easily generated from three-dimensional data by ordinary computer graphics techniques, and a person can be more easily and accurately authenticated.

According to the third aspect of the present invention, the above-described person authentication method can be implemented simply and reliably.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a person authentication device according to an embodiment of the present invention.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 ... 3D data input means 2 ... Data storage means 3 ... Position and orientation detection means 4 ... Authentication 2D image generation means 5 ... Authentication 2D image generation means 6 ...・ Image comparison means

 ──────────────────────────────────────────────────続 き Continuation of the front page (72) Inventor Yasushi Yasuda 2-13-13 Azuchicho, Chuo-ku, Osaka-shi, Osaka Inside Osaka International Building Minolta Co., Ltd. (72) Inventor Yoshitaka Isogai Azuchi-cho, Chuo-ku, Osaka-shi, Osaka 2-3-13, Osaka International Building Minolta Co., Ltd. (72) Inventor Akira Murakawa 2-3-1-13 Azuchicho, Chuo-ku, Osaka City, Osaka Prefecture Osaka International Building Minolta Co., Ltd. (72) Inventor Kenji Tsuji Osaka Osaka International Building Minolta, Inc. 2-3-1-13 Azuchicho, Chuo-ku, Osaka (72) Inventor Takashi Kondo 2-3-1-3 Azuchicho, Chuo-ku, Osaka-shi, Osaka Osaka International Building Minolta F-term ( Reference) 5B043 AA09 BA01 BA04 DA05 EA08 FA07 FA09 5B057 AA20 BA02 DC08 DC33 5L096 BA18 CA05 FA67 FA69 HA08 JA11 9A001 HZ28 HZ29 LL03

Claims (3)

[Claims] 1. Three-dimensional data of a part or the whole of a person is stored in advance as dictionary data, and a person is identified by using the dictionary data and the three-dimensional data input by a three-dimensional data input means at the time of authentication. A method of authenticating a person, comprising: detecting a position and orientation of three-dimensional data input by the three-dimensional data input means; and detecting the same position from the dictionary data and the input three-dimensional data based on the detection result. A person authentication method comprising: generating a two-dimensional image for posture authentication and a two-dimensional image for authentication; and comparing the two-dimensional image for authentication and the two-dimensional image for authentication to authenticate a person. 2. The two-dimensional image for authentication and the two-dimensional image for authentication.
The person authentication method according to claim 1, wherein a shading image or a texture mapping image is used as the two-dimensional image. 3. A three-dimensional data input means for inputting three-dimensional data on a part or the whole of a person, a storage means for storing three-dimensional data on a part or the whole of a person in advance as dictionary data, A position and orientation detecting means for detecting the position and orientation of the three-dimensional data input by the three-dimensional data input means; and the same as the dictionary data and the previously input three-dimensional data based on the detection result of the position and orientation detecting means. Two-dimensional image generating means for respectively generating a two-dimensional image for position and orientation authentication and a two-dimensional image for authentication; an image comparison for authenticating a person by comparing the two-dimensional image for authentication and the two-dimensional image for authentication with each other And a means for authenticating a person.