KR100587430B1 - System and method for photographing after correcting face pose - Google Patents

Abstract

The present invention relates to a system and method for photographing by correcting a pose of a face, comprising: a photographing unit which photographs a face image by tracking a face of a person by a pan / tilt operation and a linear motion; A database in which face images of a specific person, which have been photographed in advance at each angle, of a face pose inclined by a predetermined angle are constructed; Receiving a face image photographed by the photographing unit, comparing the input face image and the face pose inclined at a predetermined angle, the face image of a specific person built in the database, and the input image from the pre-photographed face images A fuzzy estimator configured to extract a face image having a pose corresponding to the face image; And a control unit for controlling the fuzzy estimator and receiving the signal of the fuzzy estimator.

Therefore, the present invention is to find out the inclination angle when the first face pose is inclined at a certain angle from the front face, and by moving the camera by a distance corresponding to the inclination angle so that the front face is captured by the camera lens, The front face can be photographed in real time, thereby increasing the face recognition rate.

Face, pose, correction, frontal, transfer

Description

System and method for photographing after correcting face pose}

1 is a photograph showing a front face image

2A to 2C are photographs showing a face image in which a face pose is changed by x, y, and z axes.

3 is a block diagram of a system for photographing by correcting a face pose according to the present invention;

4 is a perspective view showing a pan / tilt camera and a transfer apparatus according to the present invention;

FIG. 5 is a face photograph constructed in a database by previously photographing a face pose of a specific person who is inclined at an angle from the front face according to the present invention.

6 is a schematic perspective view of a pan / tilt camera applied to the present invention.

7A to 7D are conceptual views illustrating a concept of correcting a face pose by a system for correcting and photographing a face pose according to the present invention.

8 is a flowchart of a method of photographing by correcting a face pose according to the first embodiment of the present invention.

9 is a flowchart of a method of photographing by correcting a face pose according to a second embodiment of the present invention.

<Description of the symbols for the main parts of the drawings>

50: Pan / Tilt Camera 60: Transfer Device

100: recording unit 200: control unit

300: fuzzy estimator 400: database

The present invention relates to a system and method for photographing by correcting a face pose, and more particularly, when the first photographed face pose is tilted at a predetermined angle from the front face, the tilted angle is determined and the front face is a camera lens. The present invention relates to a system and method for correcting a face pose that can capture a front face in real time by moving a camera by a distance corresponding to an inclined angle so as to be captured.

Recently, security products using biometrics have become a bot.

In biometrics, human, unique features such as face, fingerprint, iris, retina, voice, and vein are extracted and made into a database.

In order to extract the characteristics of the user's human body, fingerprints are acquired by contact method, and iris is also troublesome to acquire an image of the eye in close proximity to the camera, and voice also requires the user to wear a microphone. have.

However, in the case of face recognition, when acquiring the face information of the user, the feature can be acquired without the user's consciousness.

Face recognition has become an important research field as an essential function for an interactive service between a user and a system in an intelligent surveillance system or a mobile robot in the future, and many methods for increasing recognition rate have been proposed.

However, in the case of face recognition, the recognition rate is sensitive depending on the surrounding environment.

FIG. 1 is a photograph showing a front face image, and FIGS. 2A to 2C are photographs showing a face image in which a face pose is converted by x, y, and z axes, and in the case of a face pose change, as shown in FIG. When an individual rotates his or her face in the x, y, z axis in the front face image, a face pose change rotated in the x, y, z axis occurs as shown in FIGS. 2A to 2C.

This change in face pose is a major factor that degrades the performance of face recognition.

In the prior art, the change to the z-axis is corrected by a simple Affine transform, but the pose is not corrected by the y-axis by the affine transform.

In another method, a change in the x-axis and the y-axis was performed by synthesizing a face image using a plurality of stereo cameras.

The use of such a plurality of stereo cameras has a problem of increasing the system cost, and also has a problem of periodically adjusting the calibration of the stereo camera, and also requires a complicated calculation for synthesis .

Another method was to construct a face image rotated on the x-axis and y-axis in the database, estimate the pose of the face image input on the eigenface, and synthesize it with software. Face-based synthesis is very difficult, and there is a problem that is not implemented in real time.

Accordingly, the present invention has been made to solve the above problems, when the first face pose is inclined at a certain angle from the front face, find out the inclination angle, and tilt the front face to be captured by the camera lens An object of the present invention is to provide a system and method for correcting and photographing a face pose that can capture a front face in real time by moving a camera by a distance corresponding to a true angle, thereby improving face recognition rate.

According to a preferred aspect of the present invention, there is provided a photographing unit which captures a face image by tracking a face of a person in a pan / tilt operation and a linear motion;

A database in which face images of a specific person, which have been photographed in advance at each angle, of a face pose inclined by a predetermined angle are constructed;

Receiving a face image photographed by the photographing unit, comparing the input face image and the face pose inclined at a predetermined angle, the face image of a specific person built in the database, and the input image from the pre-photographed face images A fuzzy estimator configured to extract a face image having a pose corresponding to the face image;

There is provided a system for controlling a fuzzy estimator and correcting a face pose including a control unit which receives a signal of the fuzzy estimator and controls the photographing unit.

Another preferred aspect for achieving the above object of the present invention is a pan / tilt camera which can move left and right, and then tracks a face of a specific person, and then photographs a face image. Obtaining a first step;

A pose corresponding to the acquired face image is obtained from the pre-photographed face images by comparing the face images constructed by photographing a face pose inclined at a predetermined angle in each front face and the acquired face image. Extracting a face image having a);

From the extracted face image, the acquired face image detects an inclination angle from the front face, moves the pan / tilt camera by a distance corresponding to the inclination angle so that the front face is captured by the camera lens, and pan / tilt operation. A method of photographing by correcting a face pose comprising a third step of performing a front face is provided.

Another preferred aspect for achieving the above object of the present invention is a pan / tilt camera that can move left and right, after tracking the face of a specific person, photographed face image Obtaining a first step;

The acquired face image is obtained by using a Principal Component Analysis (PCA) method in which face images constructed by photographing a face pose inclined at an angle from the front face in advance are inputted. Calculating a tilt angle at the face;

The camera movement distance in the database that records the camera movement distance in which the front face can be captured by the camera lens with respect to each inclination angle of the face images constructed by previously photographing a face pose inclined at a predetermined angle in the front face. Finding a third step;

And a fourth step of moving the pan / tilt camera so that the front face is captured by the camera lens by a distance corresponding to each angle inclined from the front face, and photographing the front face by performing a pan / tilt operation. A method of photographing by correcting a face pose is provided.

Hereinafter, exemplary embodiments of the present invention will be described with reference to the accompanying drawings.

3 is a configuration diagram of a system for photographing by correcting a pose of a face according to an embodiment of the present invention, wherein the photographing unit 100 captures a face image by tracking a face of a person by a pan / tilt operation and a linear motion. Wow; A database 400 in which face images of a specific person, which are photographed in advance for each angle, of a face pose inclined by a predetermined angle are constructed; The face image photographed by the photographing unit 100 is received, the face image of the specific person constructed in the database 400 is compared with the input face image and the face pose inclined at a predetermined angle, and the pre-photographed face. A fuzzy estimator 300 capable of extracting a face image having a pose corresponding to the input face image from the images; The control unit is configured to control the fuzzy estimator 300 and control the photographing unit 100 by receiving a signal from the fuzzy estimator 300.

Here, the photographing unit 100 fixes the pan / tilt camera 50 and the pan / tilt camera 50 capable of capturing a pan / tilt operation and an image, and the pan / tilt camera. It consists of a conveying apparatus 60 which can convey 50 in a straight line.

At this time, the transfer device 60 is, as shown in Figure 4, it is preferable that the transfer device that can perform the left, right linear motion using the sliding shaft (61).

In addition, the pan operation of the pan / tilt camera 50 rotates at an angle of ± 90 ° with respect to the first axis, and the tilt operation rotates at an angle of ± 45 ° with respect to the second axis.

The system for correcting and posing the facial pose according to the present invention configured as described above performs a pan / tilt operation on the pan / tilt camera 50 of the photographing unit 100 to track a specific person's face and then tracks the specific person's face. When photographing, the photographed face image is output to the fuzzy estimating unit 300, and the fuzzy estimating unit 300 is constructed by capturing a face pose of a specific person inclined at a predetermined angle from the front face for each angle in advance. The same face images are taken from the database 400, compared with the face images photographed by the pan / tilt camera 50, and the face images photographed by the pan / tilt camera 50 from the pre-photographed face images. A face image having a pose corresponding to the face is extracted.

Thereafter, the fuzzy estimator 300 transmits a signal to the controller 200 to move the pan / tilt camera by a distance corresponding to the angle of inclination of the extracted face image from the front face, and the controller The control unit 200 outputs a control signal to the transfer device 60 of the photographing unit 100.

The transfer device 60 receives the control signal from the controller 200 and moves the pan / tilt camera to capture a face of a specific person.

As a result, the system of the present invention can acquire a real-time front image with a simple configuration.

On the other hand, when the system of the present invention is mounted on a mobile robot, the mobile robot can acquire an image of ± 180 ° in order to correct the face angle, so that 100% of the face pose can be changed. By doing so, it is possible to increase the face recognition rate.

In more detail, the system of the present invention may further include a mobile robot capable of mounting the photographing unit, fuzzy estimating unit, database, and control unit.

FIG. 6 is a schematic perspective view of a pan / tilt camera applied to the present invention, and a body 2 on which the fan motor 1 is mounted, and a fan shaft 4 driven by a timing belt connected to the fan motor 1. A fan case 5 rotatably supporting one end of the fan case 5; A tilt case 6 fixedly supported at the other end of the fan shaft 4; A tilt motor 7 fixed to the side of the tilt case 6; A tilt shaft 9 for coupling one end to the tilt motor 7 and the other end rotatably fixed to the tilt case 6 to pivot the camera 8 upward and downward; The tilt shaft 9 is spaced apart at one end and is integrally fixed and consists of a bracket 10 to which the camera 8 is fixed.

When the pan / tilt camera configured as described above drives the pan motor 1 and the tilt motor 7, the camera 8 performs a pan and tilt operation.

7A to 7D are conceptual diagrams illustrating a concept of correcting a face pose by a system for correcting and photographing a face pose according to the present invention. First, as illustrated in FIG. 7A, a user may be identified as the pan / tilt camera 50. In the state where the front face of the camera can be photographed, a virtual reference line (a) in which the center of the pan / tilt camera (50) lens 51 coincides with the center of the face of a specific person is set, and this virtual reference line (a) and the face of the specific person are set. When the centers coincide with each other, the front face of the specific person may not be inclined so that the front face may be photographed.

In addition, as shown in FIG. 7B, when the face centerline of a specific person is inclined at a predetermined angle α from the aforementioned virtual reference line a, as shown in FIG. 7C, the pan / tilt camera 50 is horizontal in the system according to the present invention. After moving, as shown in FIG. 7D, a pan / tilt operation may be performed to match the center of the lens 51 of the pan / tilt camera 50 to a face of a specific person, and then the front face may be photographed.

FIG. 8 is a flowchart of a method of photographing by correcting a face pose according to a first embodiment of the present invention, and after tracking a face of a specific person with a pan / tilt camera that can move left and right. , Photographing to obtain a face image.

Thereafter, a face pose inclined at a predetermined angle from the front face is pre-photographed for each angle, and the acquired face images are compared with the acquired face images to correspond to the acquired face images from the pre-photographed face images. Extracting a face image having a pose (step 20).

From the extracted face image, the acquired face image detects an inclination angle from the front face, moves the pan / tilt camera by a distance corresponding to the inclination angle so that the front face is captured by the camera lens, and pan / tilt operation. Perform the front face to shoot (S30 step).

Here, for example, the distance to move the pan / tilt camera is built in the database as a table (Table) corresponding to each of the inclined angles, find the distance corresponding to the corresponding angle in the table, and as described above the distance The pan / tilt camera can be moved by controlling the feeder.

FIG. 9 is a flowchart of a method of photographing by correcting a pose of a face according to a second embodiment of the present invention, and after tracking a face of a specific person with a pan / tilt camera that can move left and right. , Photographing to obtain a face image.

Thereafter, the acquired face image is taken by using a Principal Component Analysis (PCA) method in which face images constructed by photographing a face pose inclined at an angle from the front face in advance at each angle are input. The angle of tilt of the image from the front face is calculated.

The camera movement distance in the database that records the camera movement distance in which the front face can be captured by the camera lens with respect to each inclination angle of the face images constructed by previously photographing a face pose inclined at a predetermined angle in the front face. Find (S300 step).

The pan / tilt camera is moved so that the front face is captured by the camera lens by a distance corresponding to each angle inclined from the front face, and the pan / tilt operation is performed to photograph the front face.

Here, in the second embodiment of the present invention, in step S200, by using a known principal component analysis (PCA) method, the angle obtained by the acquired face image from the front face is calculated, In the second embodiment, the angle is calculated using the following Equation (1).

The angle at which the acquired image is tilted at the front face = (-45 °) / (PCA output value of -45 °) + (-30 °) / (PCA output value of -30 °) + (-15 °) / (-15 PCA output value of °) + (0 °) / (PCA output value of 0 °) + (15 °) / (PCA output value of 15 °) + (30 °) / (PCA output value of 30 °) + (45 °) / (PCA output value of 45 °) ------ (1)

At this time, the Principal Component Analysis (PCA) output values of -45 °, -30 °, -15 °, 0 °, 15 °, 30 °, and 45 ° are values for the components of each angle of the obtained image.

As described above, in the present invention, when the first photographed face pose is inclined at a predetermined angle from the front face, the present invention finds out the inclination angle, and the camera corresponds to the inclination angle so that the front face is captured by the camera lens. By moving, the front face can be photographed in real time, thereby increasing the face recognition rate.                     

Although the invention has been described in detail only with respect to specific examples, it will be apparent to those skilled in the art that various modifications and variations are possible within the spirit of the invention, and such modifications and variations belong to the appended claims.

Claims (7)

A photographing unit which captures a face image by tracking a face of a person by a pan / tilt operation and a linear motion; A database in which face images of a specific person, which have been photographed in advance at each angle, of a face pose inclined by a predetermined angle are constructed; Receiving a face image photographed by the photographing unit, comparing the input face image and the face pose inclined at a predetermined angle, the face image of a specific person built in the database, and the input image from the pre-photographed face images A fuzzy estimator configured to extract a face image having a pose corresponding to the face image; And controlling a fuzzy estimator and correcting a face pose including a control unit controlling the photographing unit in response to a signal of the fuzzy estimator. The method of claim 1, The photographing unit A pan / tilt camera capable of pan / tilt operation and video, The pan / tilt camera is fixed, and the pan / tilt camera is a system for photographing by correcting the pose of the face, characterized in that it consists of a transfer device capable of transporting in a straight line. The method according to claim 1 or 2, The pan / tilt camera, The pan movement is a camera capable of rotating at an angle of ± 90 ° with respect to the first axis, and the tilt motion is a camera capable of rotating at an angle of ± 45 ° with respect to the second axis. The method of claim 1, And a photographing unit, a fuzzy estimator, a database, and a mobile robot equipped with a control unit. A first step of tracking a face of a specific person with a pan / tilt camera capable of moving left and right, and then capturing a face to obtain a face image; A pose corresponding to the acquired face image is obtained from the pre-photographed face images by comparing the face images constructed by photographing a face pose inclined at a predetermined angle in each front face and the acquired face image. Extracting a face image having a); From the extracted face image, the acquired face image detects an inclination angle from the front face, moves the pan / tilt camera by a distance corresponding to the inclination angle so that the front face is captured by the camera lens, and pan / tilt operation. Performing a face pose consisting of a third step of photographing the front face. A first step of tracking a face of a specific person with a pan / tilt camera capable of moving left and right, and then capturing a face to obtain a face image; The acquired face image is obtained by using a Principal Component Analysis (PCA) method in which face images constructed by photographing a face pose inclined at an angle from the front face in advance are input. Calculating a tilt angle at the face; A third step of finding the camera movement distance in a database in which the camera movement distance in which the front face can be captured by the camera lens with respect to the angle calculated in the second step; A method of correcting and photographing a face pose comprising a fourth step of moving the pan / tilt camera so that the front face is captured by the camera lens by the camera movement distance found in the third step, and performing a pan / tilt operation to photograph the front face . The method of claim 6, In the second step, using the Principal Component Analysis (PCA) method, calculating the angle of inclination of the acquired face image in the front face is a face pose characterized in that the following equation (1) How to calibrate and shoot. The angle at which the acquired image is tilted at the front face = (-45 °) / (PCA output value of -45 °) + (-30 °) / (PCA output value of -30 °) + (-15 °) / (-15 PCA output value of °) + (0 °) / (PCA output value of 0 °) + (15 °) / (PCA output value of 15 °) + (30 °) / (PCA output value of 30 °) + (45 °) / (PCA output value of 45 °) ------ (1)

Images