KR20210129398A - Face recognition device using deep learning - Google Patents

Abstract

According to an embodiment of the present invention, a face recognition device includes: a reference image input unit for receiving a reference image; a region of interest setting unit for generating an image of interest by setting a region of interest corresponding to the reference image; a learning image generating unit for generating a plurality of learning images based on the image of interest; a learning unit for generating a recognition model by learning using the plurality of learning images; and a face recognition unit for recognizing a face from an input image using the recognition model. Accordingly, a recognition rate can be improved in various recognition situations.

Description

Face recognition device using deep learning {FACE RECOGNITION DEVICE USING DEEP LEARNING}

An embodiment of the present invention relates to a face recognition apparatus using deep learning, in particular, a face recognition apparatus capable of generating a plurality of learning images using a reference image and learning the plurality of learning images to improve the recognition rate in various recognition situations. will be.

Face detection refers to a technology for detecting a human face from an input image.

Face recognition refers to a technology that detects a human face from an input image and finds a specific face or person by comparing it with faces stored in a database (DB).

Recently, face detection and face recognition have been widely used in various fields. For example, in the field of a security system for controlling access to a security system or in the field of marketing that provides appropriate marketing services according to people to potential customers, face detection and face recognition systems are being used.

Face detection and face recognition are being implemented using not only separate cameras such as CCTV, but also mobile phones, smartphones, tablet PCs, and laptops.

The conventional face recognition technology has a problem in that real-time processing is difficult because the amount of data to be processed and the amount of computation are huge.

In addition, in various face recognition situations (eg, a situation in which illumination is low, a situation in which there is an expression, a situation in which a light of a specific color illuminates a face, a situation in which the camera is not viewed directly, etc.), the conventional face recognition technology has a face recognition rate There is a limit to this falling.

And, the resolution of the image used for face recognition is increasing day by day. For example, the input image has a high resolution such as SD, HD, FULL HD, 4K, etc., and the capacity thereof is increasing. Therefore, for real-time processing of face recognition, more computing power is required.

Korean Patent No. 10-2016082 'Deep learning-based facial recognition method and device robust to pose change' (published on Aug. 23, 2019) Korean Patent Registration No. 10-1962875 'A device and method for generating features for face recognition in low-resolution images' (published on June 30, 2017)

SUMMARY OF THE INVENTION It is an object of the present invention to provide a face recognition apparatus capable of improving a recognition rate in various recognition situations.

Another object of the present invention is to provide a face recognition device that requires less computing power even when an image of the same size is input.

Another object of the present invention is to provide a face recognition apparatus capable of accurately recognizing a face only with a side image of a human face.

According to an embodiment of the present invention, a face recognition apparatus includes: a reference image input unit for receiving a reference image; a region of interest setting unit for generating an image of interest by setting a region of interest corresponding to the reference image; a learning image generator for generating a plurality of learning images based on the image of interest; a learning unit for generating a recognition model by learning using the plurality of training images; and a face recognition unit for recognizing a face from an input image by using the recognition model.

In the present invention, the learning image generating unit is characterized in that it includes an angle changing unit for generating a plurality of learning images by rotating the angle of the image of interest at least once according to a unit angle.

In the present invention, the unit angle is characterized in that it belongs to the range of 5 degrees to 15 degrees.

In the present invention, the training image generating unit is characterized in that it further comprises a brightness changing unit for generating a plurality of training images by setting the brightness of the image of interest according to the unit brightness.

In the present invention, the brightness changing unit generates the plurality of training images according to the unit brightness in a range of +100% to -100% based on the brightness of the image of interest, and the unit brightness is 5% to 10% It is characterized in that it falls within the range.

In the present invention, the training image generator may further include a contrast changer configured to generate a plurality of training images by setting the contrast of the image of interest according to the unit intensity.

In the present invention, the contrast change unit generates a plurality of training images according to the unit contrast in a range of +100% to -100% based on the contrast of the image of interest, and the unit contrast is in the range of 5% to 10% It is characterized by belonging to

In the present invention, the training image generator may further include a color changer configured to generate a plurality of training images by extracting the color of the image of interest according to a plurality of unit color channels.

In the present invention, the plurality of unit color channels includes a red channel, a green channel, and a blue channel.

In the present invention, the plurality of unit color channels includes a cyan channel, a magenta channel, a yellow channel, and a black channel.

In the present invention, the learning image generating unit may further include a sharpness adjusting unit for generating a plurality of learning images by adjusting the sharpness of the outline of the image of interest.

In the present invention, the learning image generating unit, when the image of interest represents the side part of the face, characterized in that it further comprises a front image generating unit for generating a front image through left and right inversion and image combination.

In the present invention, the front image generating unit provides the front image to the angle changing unit, the brightness changing unit, the contrast changing unit, the color changing unit, and the sharpness adjusting unit.

In the present invention, the ROI setting unit generates the ROI by detecting a shape of a face or a person from the reference image.

In the present invention, the learning unit, a filter unit for removing images having a reference resolution or less from among the plurality of training images; an extraction unit for extracting feature points of the plurality of learning images; and a model manager for generating the recognition model based on the feature point.

The face recognition apparatus of the present invention has an effect of improving the recognition rate in various recognition situations.

In addition, the face recognition apparatus of the present invention has the effect of requiring less computing power even when an image of the same size is input.

In addition, the face recognition apparatus of the present invention has an effect of accurately recognizing a face only with a side image of a human face.

1 is a diagram illustrating a face recognition apparatus according to an embodiment of the present invention.
2 is a diagram illustrating an operation of a region of interest setting unit according to an embodiment of the present invention.
3 is a diagram illustrating a learning image generator according to an embodiment of the present invention.
4 is a diagram illustrating an operation of an angle changing unit according to an embodiment of the present invention.
5 is a diagram illustrating an operation of a front image generator according to an embodiment of the present invention.
6 is a diagram illustrating a learning unit according to an embodiment of the present invention.

The present invention will be described in more detail.

Hereinafter, embodiments of the present invention and other matters necessary for those skilled in the art to easily understand the contents of the present invention will be described in detail with reference to the accompanying drawings. However, since the present invention can be embodied in various different forms within the scope of the claims, the embodiments described below are merely exemplary regardless of whether they are expressed or not.

Like reference numerals refer to like elements. In addition, in the drawings, thicknesses, ratios, and dimensions of components are exaggerated for effective description of technical content. “And/or” may include any combination of one or more that the associated configurations may define.

Terms such as first, second, etc. may be used to describe various elements, but the elements should not be limited by the terms. The above terms are used only for the purpose of distinguishing one component from another. For example, without departing from the scope of the present invention, a first component may be referred to as a second component, and similarly, a second component may also be referred to as a first component. The singular expression may include the plural expression unless the context clearly dictates otherwise.

In addition, terms such as "below", "below", "above", "upper" and the like are used to describe the relationship between the components shown in the drawings. The above terms are relative concepts, and are described based on directions indicated in the drawings.

Terms such as “comprise” or “have” are intended to designate that a feature, number, step, action, component, part, or combination thereof described in the specification is present, and includes one or more other features, numbers, or steps. , it should be understood that it does not preclude the possibility of the existence or addition of , operation, components, parts, or combinations thereof.

That is, the present invention is not limited to the embodiments disclosed below, but may be implemented in a variety of different forms, and in the following description, when a part is connected to another part, it is directly connected In addition, it may include a case in which another element is electrically connected therebetween. In addition, it should be noted that the same elements in the drawings are denoted by the same reference numbers and symbols as much as possible even though they are indicated in different drawings.

1 is a diagram illustrating a face recognition apparatus 10 according to an embodiment of the present invention.

Referring to FIG. 1 , the face recognition apparatus 10 includes a reference image input unit 100 , a region of interest setting unit 200 , a training image generating unit 300 , a learning unit 400 , and a face recognition unit 500 . may include

The reference image input unit 100 may receive a reference image. For example, the reference image may be an image photographed by a separate photographing device or an image stored in a storage medium. The reference image may be an image obtained by photographing a human face or a human body.

The region of interest setting unit 200 may generate an image of interest by setting a region of interest corresponding to the reference image. For example, the region of interest setting unit 200 may detect a shape of a face or a person from the reference image. The region of interest setting unit 200 may generate a separate image of interest by setting coordinates of the face region and cropping an image corresponding to the set coordinates. In this case, the image of interest may represent a human face.

The training image generator 300 may generate a plurality of training images based on the image of interest. For example, the training image generator 300 may generate a plurality of training images by regenerating images in various situations (eg, lighting, color, sharpness, brightness, contrast, etc.) based on the image of interest. In addition, when the image of interest represents a side part of a human face, the learning image generating unit 300 calculates a front image by using image inversion and image combining technology, and using the front image, it is possible to generate a plurality of learning images. have.

The learning unit 400 may generate a recognition model by learning using a plurality of learning images. For example, the learning unit 400 may generate a recognition model capable of recognizing a human face in various situations by using a plurality of training images generated by the training image generator.

According to an embodiment, the learning unit 400 may be implemented as a learning model such as a Convolutional Neural Network (CNN) or a Recurrent Neural Network (RNN). CNN is a learning model mainly used to recognize images, and even for the same image, there may be various cases such as a change of position, a change of direction, or an image distortion. It refers to a learning model that extracts abstracted information from images by repeatedly applying and pooling processes. Convolution refers to a method of deriving image features using various filters, and pooling refers to a method of reducing dimensions while maintaining image features. RNN is a learning model used to recognize sequential information, and is mainly used for speech or character recognition.

The face recognition unit 500 may recognize a face from an input image by using the recognition model generated by the learning unit.

For example, the face recognition unit 500 may recognize a face using an appearance based method. The shape-based method expresses an image as a two-dimensional array or vector and uses statistical information of brightness values of pixels to recognize a face. Eigenface, Fisherface, Discriminant Common Vector (DCV); discriminant common vectors), and various methods that extend and transform them. For example, various other methods include Direct Linear Discriminant Analysis (LDA), Eigenfeature Regularization and Extraction (ERE), and Marginal Fisher Analysis (MFA). The present invention is not limited thereto, and according to embodiments, the face recognition unit 500 may recognize a face using various face recognition methods.

2 is a diagram illustrating an operation of the region of interest setting unit 200 according to an embodiment of the present invention.

As shown in FIG. 2 , the region of interest setting unit 200 may generate an image of interest IIM by setting a region of interest corresponding to the reference image RIM. For example, the region of interest setting unit 200 may detect a shape of a face or a person from the reference image RIM. That is, the region of interest setting unit 200 may detect the face region FC. The region of interest setting unit 200 may set the coordinates of the face region FC, and may generate a separate image of interest IIM by cropping an image corresponding to the set coordinates. In this case, the image of interest may represent a human face region FC.

According to an embodiment, the region of interest setting unit 200 may detect the face region FC from the reference image RIM by using a deep learning algorithm. The deep learning algorithm is a kind of machine learning technique, and refers to an artificial neural network algorithm including a plurality of hidden layers. In order to use the deep learning algorithm, the deep learning algorithm model must be preceded by a learning step using images including various face poses, and the following description assumes that the deep learning algorithm mentioned in this specification has completed this learning step. . According to an embodiment, the region of interest setting unit 200 may recognize a face using the recognition model generated by the learning unit 400 shown in FIG. 1 .

The region of interest setting unit 200 may detect the face region FC from the reference image RIM and generate face region information including coordinates regarding the detected face region. Here, the face region information may include at least one of x-coordinate, y-coordinate, width, and height information of the face region FC in the reference image RIM. In this case, the x-coordinate and the y-coordinate may be coordinates of the center point of the face region FC or coordinates of all image pixels.

According to an embodiment, the region of interest setting unit 200 may generate the image of interest IIM having various shapes (eg, a rectangle, a triangle, a circle, an oval, a semicircle, etc.) according to the face region FC.

Also, the region of interest setting unit 200 may normalize the image of interest IIM. For example, the region of interest setting unit 200 may perform a normalization process by setting the image of interest IIM to an image having a preset size or by evening color distribution.

3 is a diagram illustrating a learning image generating unit 300 according to an embodiment of the present invention.

Referring to FIG. 3 , the learning image generating unit 300 includes an angle change unit 310 , a brightness change unit 320 , a contrast change unit 330 , a color change unit 340 , a sharpness adjuster 350 , and a front surface. It may include an image generator 360 .

The angle changer 310 may generate a plurality of learning images by rotating the angle of the image of interest at least once according to a unit angle.

For example, the angle changer 310 may generate a plurality of learning images by rotating the image of interest at least once by a unit angle in a clockwise or counterclockwise direction. According to an embodiment, the unit angle may be in a range of 5 degrees to 15 degrees.

According to an embodiment, the angle changer 310 may recognize a pose of a person in the image of interest. Then, the angle changer 310 expands the image of interest in three dimensions based on the pose of the image of interest, and by a unit angle along at least one axis of yaw, pitch, and roll. A plurality of learning images may be generated by rotating at least once or more.

The brightness changer 320 may generate a plurality of learning images by setting the brightness of the image of interest according to the unit brightness. For example, the brightness changing unit 320 may generate a plurality of learning images by setting the brightness of the image of interest in the range of +100% to -100% according to the unit brightness based on the brightness. In this case, the unit brightness may be in the range of 5% to 10%. Through this, the face recognition apparatus 10 according to an embodiment of the present invention may improve the face recognition rate in various lighting situations.

The contrast changer 330 may generate a plurality of learning images by setting the contrast of the imaging of interest according to the unit contrast. For example, the contrast changer 330 may generate a plurality of learning images by setting the intensity of the image of interest in the range of +100% to -100% according to the unit intensity based on the intensity. In this case, the unit contrast may be in the range of 5% to 10%. Through this, the face recognition apparatus 10 according to an embodiment of the present invention may improve the face recognition rate in various lighting situations.

According to an embodiment, the contrast changing unit 330 may maximize the contrast by changing the contrast. That is, the contrast changer 330 may maximize the contrast by dividing the image of interest into a plurality of regions and adjusting the average brightness and contrast for each region.

The color changer 340 may generate a plurality of learning images by extracting colors of the image of interest according to a plurality of unit color channels. For example, the plurality of unit color channels may include a red channel, a green channel, and a blue channel. However, the present invention is not limited thereto, and according to embodiments, the plurality of unit color channels may include a cyan channel, a magenta channel, a yellow channel, and a black channel. Through this, the face recognition apparatus 10 according to an embodiment of the present invention may improve the face recognition rate in various lighting color situations.

The sharpness adjusting unit 350 may generate a plurality of learning images by adjusting the sharpness of the outline of the image of interest. For example, the sharpness adjusting unit 350 may detect an outline of the image of interest and adjust the sharpness so that the outline is more clearly revealed.

The front image generator 360 may generate a front image through left-right inversion and image combination when the image of interest represents the side part of the face. In addition, the front image generator 360 provides the front image to the angle changer 310 , the brightness changer 320 , the contrast changer 330 , the color changer 340 , and the sharpness adjuster 350 . can Accordingly, each of the angle changer 310, the brightness changer 320, the contrast changer 330, the color changer 340, and the sharpness adjuster 350 may perform the above-described operation on the image of interest. have.

For example, when the image of interest represents the right side of the face, the front image generator 360 may generate an image representing the left side by inverting the image of interest, and may generate the front image by combining the two images. Also, the front image generator 360 may perform 3D modeling based on the image of interest and extract the front image through this.

4 is a diagram illustrating an operation of the angle changing unit 310 according to an embodiment of the present invention.

Referring to FIG. 4 , the angle changer 310 may generate a plurality of learning images LIM by rotating an image of interest at least once according to a unit angle.

For convenience of explanation, FIG. 4 shows the contents of the image of interest rotated by 90 degrees in the clockwise direction, but the present invention is not limited thereto. According to an embodiment, the image of interest may be rotated 360 degrees.

The angle changer 310 may generate a plurality of learning images LIM by rotating the image of interest at least once by a unit angle in the clockwise direction. The unit angle may be in the range of 5 degrees to 15 degrees. Through this, the face recognition apparatus according to an embodiment of the present invention may improve the face recognition rate by performing learning using images rotated at various angles and performing face recognition using the model derived accordingly.

According to an embodiment, the angle changer 310 may recognize a posture of a person from the image of interest. In addition, the angle change unit 310 expands the image of interest in three dimensions based on the posture of the person, and along at least one axis of yaw, pitch, and roll, as shown in FIG. A plurality of learning images may be generated by rotating at least one time or more by a unit angle similarly as illustrated.

5 is a diagram illustrating an operation of the front image generator 360 according to an embodiment of the present invention.

Referring to FIG. 5 , when the image of interest represents the side part of the face, the front image generator 360 may generate the front image IIM′ through left-right inversion and image combination. In addition, the front image generator 360 converts the front image IIM' into the angle changer 310 , the brightness changer 320 , the contrast changer 330 , the color changer 340 , and the sharpness adjuster 350 . ) can be provided. Accordingly, each of the angle changer 310, the brightness changer 320, the contrast changer 330, the color changer 340, and the sharpness adjuster 350 performs the above-described operation for the image of interest in the front image ( IIM').

For example, when the image of interest represents the left side of the face, the front image generator 360 may generate the right side image IIMR by inverting the left side image IIML based on the left side. In addition, the front image generator 360 may generate the front image IIM' by combining the left side image IIML and the right side image IIMR.

According to an embodiment, the front image generator 360 may perform 3D modeling based on the image of interest and extract the front image through this.

6 is a diagram illustrating a learning unit 400 according to an embodiment of the present invention.

Referring to FIG. 6 , the learning unit 400 may include a filter unit 410 , an extraction unit 420 , and a model management unit 430 .

The filter unit 410 may remove images having a reference resolution or less from among the plurality of training images. For example, the filter unit 410 may remove an image having a resolution lower than or equal to a reference resolution in order to improve learning accuracy and efficiency. According to an embodiment, the reference resolution may be set to various values and may be changed according to a user input. In this case, the reference resolution may indicate that the number of pixels on a horizontal or vertical axis is 200. Alternatively, the reference resolution may mean a relative ratio of horizontal and vertical. That is, when the resolution is excessively low or the ratio of the image is incorrect, the filter unit 410 may remove the corresponding image to increase learning efficiency.

The extractor 420 may extract feature points of a plurality of learning images. For example, the extraction unit 420 may extract a feature point from the plurality of training images filtered by the filter unit 410 . According to an embodiment, various known methods related to extracting feature points of an image may be utilized.

The model manager 430 may generate a recognition model based on the feature points. For example, the model manager 430 may generate and manage a recognition model by performing learning based on the feature points.

Through the above-described method, the face recognition apparatus of the present invention has an effect of improving the recognition rate in various recognition situations.

In addition, the face recognition apparatus of the present invention has the effect of requiring less computing power even when an image of the same size is input.

In addition, the face recognition apparatus of the present invention has an effect of accurately recognizing a face only with a side image of a human face.

The functional operations described herein and the embodiments related to the present subject matter are implemented in digital electronic circuits or computer software, firmware or hardware, including the structures disclosed herein and structural equivalents thereof, or in a combination of one or more of these possible.

Embodiments of the subject matter described herein relate to one or more computer program products, ie one or more computer program instructions encoded on a tangible program medium for execution by or for controlling the operation of a data processing device. It can be implemented as a module. A tangible program medium may be a radio wave signal or a computer-readable medium. A radio wave signal is an artificially generated signal, such as, for example, a machine-generated electrical, optical or electromagnetic signal, that is generated to encode information for transmission to an appropriate receiver device for execution by a computer. The computer-readable medium may be a machine-readable storage device, a machine-readable storage substrate, a memory device, a combination of materials that affect a machine-readable radio wave signal, or a combination of one or more of these.

A computer program (also known as a program, software, software application, script or code) may be written in any form of any programming language, including compiled or interpreted language or a priori or procedural language, as a stand-alone program or module; It can be deployed in any form, including components, subroutines, or other units suitable for use in a computer environment.

A computer program does not necessarily correspond to a file on a file device. A program may be in a single file provided to the requested program, or in multiple interacting files (eg, files storing one or more modules, subprograms, or portions of code), or in files holding other programs or data. It may be stored within some (eg, one or more scripts stored within a markup language document).

The computer program may be deployed to be executed on a single computer or multiple computers located at one site or distributed over a plurality of sites and interconnected by a communication network.

Additionally, the logic flows and structural block diagrams described in this patent document describe corresponding acts and/or specific methods supported by corresponding functions and steps supported by the disclosed structural means, and corresponding It can also be used to build software structures and algorithms and their equivalents.

The processes and logic flows described herein may be performed by one or more programmable processors executing one or more computer programs to perform functions by operating on input data and generating outputs.

Processors suitable for the execution of computer programs include, for example, both general and special purpose microprocessors and any one or more processors of any form of digital computer. Typically, the processor will receive instructions and data from read-only memory or random access memory or both.

A key component of a computer is one or more memory devices for storing instructions and data and a processor for executing instructions. In addition, a computer is generally configured to be operable to receive data from, transfer data to, or perform both operations on one or more mass storage devices for storing data, such as, for example, magnetic, magneto-optical disks or optical disks. combined or will include. However, the computer need not have such a device.

The present description sets forth the best mode of the invention, and provides examples to illustrate the invention, and to enable any person skilled in the art to make or use the invention. This written specification does not limit the present invention to the specific terms presented.

Although the above has been described with reference to the preferred embodiment of the present invention, those skilled in the art or those having ordinary knowledge in the technical field will not depart from the spirit and technical scope of the present invention described in the claims to be described later. It will be understood that various modifications and variations of the present invention can be made without departing from the scope of the present invention.

Accordingly, the technical scope of the present invention should not be limited to the contents described in the detailed description of the specification, but should be defined by the claims.

10: face recognition device 100: reference image input unit
200: region of interest setting unit 300: learning image generating unit
310: angle change unit 320: brightness change unit
330: contrast change unit 340: color change unit
350: sharpness control unit 360: front image generation unit
400: learning unit 410: filter unit
420: extraction unit 430: model management unit
500: face recognition unit

Claims (15)

a reference image input unit for receiving a reference image;
a region of interest setting unit for generating an image of interest by setting a region of interest corresponding to the reference image;
a learning image generator for generating a plurality of learning images based on the image of interest;
a learning unit for generating a recognition model by learning using the plurality of training images; and
Using the recognition model, characterized in that it comprises a face recognition unit for recognizing a face from the input image,
face recognition device. According to claim 1,
The learning image generation unit,
It characterized in that it comprises an angle changer for generating a plurality of learning images by rotating the angle of the image of interest at least once according to the unit angle,
face recognition device. 3. The method of claim 2,
The unit angle, characterized in that it belongs to the range of 5 degrees to 15 degrees,
face recognition device. 3. The method of claim 2,
The learning image generation unit,
Characterized in that it further comprises a brightness changing unit for generating a plurality of learning images by setting the brightness of the image of interest according to the unit brightness,
face recognition device. 5. The method of claim 4,
The brightness changing unit generates the plurality of learning images according to the unit brightness in a range of +100% to -100% based on the brightness of the image of interest,
The unit brightness is characterized in that it belongs to the range of 5% to 10%,
face recognition device. 5. The method of claim 4,
The learning image generation unit,
Characterized in that it further comprises a contrast change unit for generating a plurality of learning images by setting the contrast of the image of interest according to the unit contrast,
face recognition device. 7. The method of claim 6,
The contrast change unit generates a plurality of learning images according to the unit contrast in a range of +100% to -100% based on the contrast of the image of interest,
The unit contrast, characterized in that it belongs to the range of 5% to 10%,
face recognition device. 7. The method of claim 6,
The learning image generation unit,
Characterized in that it further comprises a color changer for generating a plurality of learning images by extracting the color of the image of interest according to a plurality of unit color channels,
face recognition device. 9. The method of claim 8,
wherein the plurality of unit color channels include a red channel, a green channel, and a blue channel,
face recognition device. 9. The method of claim 8,
wherein the plurality of unit color channels include a cyan channel, a magenta channel, a yellow channel, and a black channel.
face recognition device. 9. The method of claim 8,
The learning image generation unit,
Characterized in that it further comprises a sharpness adjusting unit for generating a plurality of learning images by adjusting the sharpness of the outline of the image of interest,
face recognition device. 12. The method of claim 11,
The learning image generation unit,
When the image of interest indicates the side part of the face, characterized in that it further comprises a front image generating unit for generating a front image through left and right inversion and image combination,
face recognition device. 13. The method of claim 12,
wherein the front image generating unit provides the front image to the angle changing unit, the brightness changing unit, the contrast changing unit, the color changing unit and the sharpness adjusting unit,
face recognition device. According to claim 1,
The region of interest setting unit, characterized in that by detecting the shape of a face or a person in the reference image, to generate the image of interest,
face recognition device. 15. The method of claim 14,
The learning unit,
a filter unit for removing images of a reference resolution or less from among the plurality of training images;
an extraction unit for extracting feature points of the plurality of learning images; and
Characterized in that it comprises a model management unit for generating the recognition model based on the feature point,
face recognition device.

Images