KR102367154B1 - Face identification method, and device and program using the same - Google Patents

Abstract

The face identification method according to an embodiment of the present invention includes detecting a face of an identification target included in at least one image frame among the plurality of image frames at each predetermined identification period with respect to a plurality of collected image frames. , performing face identification on the image frame in which the face of the subject to be identified is detected, and adjusting the identification cycle when the face of the subject to be identified is detected but not identified according to the result of performing the face identification include

Description

FACE IDENTIFICATION METHOD, AND DEVICE AND PROGRAM USING THE SAME

The present invention relates to a face identification method, and an apparatus and a program using the same, and more particularly, a face identification method capable of adjusting an identification cycle for identifying a face of an identification target when the identification target's face is detected but not identified; And it relates to an apparatus and a program using the same.

There are various algorithms such as K-means clustering, K-Nearest Neighbors (KNN) algorithm, and Support Vector Machine (SVM) algorithm for data classification.

Most of the algorithms for data classification can perform clustering on data based on the distance between the data. can be classified.

An object of the present invention is to provide a face identification method capable of adjusting an identification cycle for recognizing an identification target's face when the identification target's face is detected but not identified, and an apparatus and program using the same.

The face identification method according to an embodiment of the present invention includes detecting the face of an identification target included in at least one image frame among the plurality of image frames at each predetermined identification period with respect to a plurality of collected image frames. , performing face identification on the image frame in which the face of the subject to be identified is detected, and adjusting the identification cycle when the face of the subject to be identified is detected but not identified according to the result of performing the face identification may include

According to an embodiment, the adjusting of the identification period may maintain the identification period when the face of the identification target is not detected in the at least one image frame.

According to an embodiment, the adjusting of the identification period may reduce the identification period when the face of the identification target is detected but not identified according to a result of performing the face identification.

According to an embodiment, the adjusting of the identification period may include, in a case in which a face is not identified according to a result of performing the face identification, first among when the face is identified and when a preset first time limit has elapsed. The reduced identification period can be maintained until it arrives.

According to an embodiment, the performing of the face identification may include generating an N-dimensional vector (where N is a natural number equal to or greater than 2) corresponding to the face image from a face image including the face of the identified subject; Calculating a distance value between an N-dimensional vector and each of a plurality of pre-stored N-dimensional vectors and N in which the distance value between vectors calculated from among a plurality of pre-stored N-dimensional vectors for the identification target is within a reference distance value identifying the person corresponding to the dimension vector.

According to an embodiment, the face identification method includes adjusting the time limit based on a distance between an N-dimensional vector corresponding to the identification target and the nearest N-dimensional vector among the plurality of pre-stored N-dimensional vectors may include more.

According to an embodiment, the adjusting of the time limit may include adjusting the first time limit to be longer as the distance between the nearest N-dimensional vectors is relatively close.

According to an embodiment, in the face identification method, when the face of the identification target is not identified until the preset first time limit elapses, the reference distance value is reduced during a second time limit or the second limit The collection of image frames can be stopped for a period of time.

The face identification apparatus according to an embodiment of the present invention detects a face of a subject to be identified included in at least one image frame among the plurality of image frames at each predetermined identification period with respect to a plurality of collected image frames. A module, a face identification module for performing face identification with respect to an image frame in which a face of a subject to be identified is detected, and a parameter adjusting module for adjusting the identification cycle when a face is not identified according to a result of performing the face identification can do.

A program stored in a medium for performing the face identification method in combination with a processor according to an embodiment of the present invention is at least one of the plurality of image frames for each predetermined identification period for a plurality of collected image frames. detecting the face of the subject to be identified included in the image frame of It may include a program code for performing the step of adjusting the identification period when not identified.

Methods and apparatuses according to an embodiment of the present invention identify the face of the subject to be identified at regular identification intervals, but adjust the identification period for identifying the face of the subject to be identified when the face of the subject to be identified is detected but not identified. It is possible to reduce the processing load by performing the identification of .

In addition, the method and apparatus according to an embodiment of the present invention reduce the identification period for recognizing the face of the subject to be identified when the face of the subject to be identified is detected but not identified, thereby intensively performing the identification process to reduce motion blur, etc. There is an effect that can significantly reduce the instantaneous identification error caused by

In order to more fully understand the drawings recited in the Detailed Description, a brief description of each drawing is provided.
1 is a block diagram of a face identification apparatus according to an embodiment of the present invention.
FIG. 2 is a diagram illustrating N-dimensional vectors used in the face identification apparatus of FIG. 1 .
FIG. 3 is a diagram for explaining a process of calculating a distance between N-dimensional vectors illustrated in FIG. 2 .
4 is a flowchart of a face identification method according to an embodiment of the present invention.
5 is a diagram illustrating a process of performing face identification according to the face identification method of FIG. 4 .

Since the technical spirit of the present invention can have various changes and can have various embodiments, specific embodiments are illustrated in the drawings and described in detail. However, this is not intended to limit the technical spirit of the present invention to specific embodiments, and it should be understood to include all changes, equivalents, or substitutes included in the scope of the technical spirit of the present invention.

In describing the technical idea of the present invention, if it is determined that a detailed description of a related known technology may unnecessarily obscure the subject matter of the present invention, the detailed description thereof will be omitted. In addition, numbers (eg, first, second, etc.) used in the description process of the present specification are only identification symbols for distinguishing one component from other components.

In addition, in this specification, when a component is referred to as “connected” or “connected” with another component, the component may be directly connected or directly connected to the other component, but in particular It should be understood that, unless there is a description to the contrary, it may be connected or connected through another element in the middle.

In addition, terms such as "~ unit", "~ group", "~ character", and "~ module" described in this specification mean a unit that processes at least one function or operation, which is a processor, a micro Processor (Micro Processor), Micro Controller (Micro Controller), CPU (Central Processing Unit), GPU (Graphics Processing Unit), APU (Accelerate Processor Unit), DSP (Drive Signal Processor), ASIC (Application Specific Integrated Circuit), FPGA (Field Programmable Gate Array), etc. may be implemented as hardware or software or a combination of hardware and software, and may be implemented in a form combined with a memory that stores data necessary for processing at least one function or operation. .

In addition, it is intended to clarify that the classification of the constituent parts in the present specification is merely a classification for each main function that each constituent unit is responsible for. That is, two or more components to be described below may be combined into one component, or one component may be divided into two or more for each more subdivided function. In addition, each of the constituent units to be described below may additionally perform some or all of the functions of other constituent units in addition to the main function it is responsible for. Of course, it may be carried out by being dedicated to it.

1 is a block diagram of a face identification apparatus according to an embodiment of the present invention. FIG. 2 is a diagram illustrating N-dimensional vectors used in the face identification apparatus of FIG. 1 . FIG. 3 is a diagram for explaining a process of calculating a distance between the N-dimensional vectors shown in FIG. 2 .

Referring to FIG. 1 , the face identification neglect 100 may include an image capture unit 110 , a memory 120 , a communication interface 130 , and a processor 140 .

According to an embodiment, the face identification apparatus 100 may be implemented as various types of devices installed in a vehicle to register, identify, and authenticate the driver's face, and the application field is not limited thereto.

According to an embodiment, the image capture unit 110 may be implemented separately from the face identification apparatus 100 , and in this case, the face identification apparatus 100 receives image data from the separately implemented image capture unit 110 . can receive

The image capture unit 110 may capture and obtain image frames including a human face.

In this specification, the term “image” may refer to a concept including both a static image and a dynamic image (ie, an image).

According to an embodiment, the image capture unit 110 may include an optical system including an image sensor (eg, a charge coupled device (CCD), a complementary metal-oxide semiconductor (CMOS), etc.) for capturing an image and a lens. .

The memory 120 may store data or a program necessary for the overall operation of the traffic signal identification apparatus 100 .

The memory 120 may store an image acquired by the image capture unit 110 or an image being processed or processed by the processor 140 .

The communication interface 130 may interface communication between the traffic signal identification apparatus 100 and another external device.

For example, the communication interface 130 may transmit information about the subject identified by the face identification device 100 to the vehicle control system, the driving assistance control system, or the autonomous driving control system.

The processor 140 may include an image preprocessing module 142 , a face detection module 144 , a face identification module 146 , and a parameter adjustment module 148 .

In this specification, a module may mean a functional or structural combination of hardware for performing a method according to an embodiment of the present invention or software capable of driving the hardware.

Accordingly, the module may mean a logical unit or set of program code and a hardware resource capable of executing the program code, and does not necessarily mean physically connected code or one type of hardware.

It indicates that each of the image pre-processing module 142, the face detection module 144, the face identification module 146, and the parameter adjustment module 148 can be functionally and logically separated, and each of the components must be a separate physical It is not meant to be device-specific or written in separate code.

The image pre-processing module 142 may perform a pre-processing process on the image frames acquired by the image capture unit 110 .

According to an embodiment, the image pre-processing module 142 may perform noise removal, gamma correction, color filter array interpolation, color correction, Image processing for image quality improvement such as color enhancement may be performed.

The face detection module 144 may detect the face of the identification target included in at least one image frame among the image frames processed by the image pre-processing module 142 .

According to an embodiment, the face detection module 144 may detect the face of the subject included in the acquired image. For example, the image preprocessing module 142 uses an algorithm such as Scale Invariant Feature Transform (SIFT), Histogram of Oriented Gradient (HOG), Haar, Ferns, Local Binary Pattern (LBP), or Modified Census Transform (MCT). faces can be detected.

According to an embodiment, the face detection module 144 may correct a face bounding box, which is a region set for detecting the subject's face, in order to accurately detect the face of the subject to be identified within the acquired image frame. . For example, the face bounding box may be set differently depending on the type of algorithm used for face detection, the face shape of the subject, or lighting.

According to an embodiment, the face detection module 144 may detect a feature point, a feature vector, or a landmark from the image frame.

According to an embodiment, the face detection module 144 may align the faces of the subjects detected in the image frame. For example, the face detection module 144 may align the faces in the image frame with reference coordinates by using a feature point, a feature vector, or a landmark detected in the acquired image frame.

According to an embodiment, the face detection module 144 may generate an N-dimensional vector (where N is a natural number equal to or greater than 2) corresponding to the face of the identification target.

According to an embodiment, the N-dimensional vector may mean a vector composed of N natural numbers.

According to an embodiment, the face detection module 144 may detect the face of the identification target included in at least one image frame among a plurality of image frames at every preset identification cycle.

The face identification module 146 may perform face identification on an image frame in which the face of the subject to be identified is detected by the face detection module 144 .

The face identification module 146 may perform face identification by comparing an N-dimensional vector corresponding to the detected face with a plurality of N-dimensional vectors pre-stored in the memory 120 with respect to the image frame in which the face of the person to be identified is detected. there is.

2 and 3 together, N-dimensional vectors NV-OG1 to NV-OGM corresponding to faces of authenticated people may be pre-stored in the memory 120 .

The face detection module 144 may generate an N-dimensional vector (eg, NV-NEW1 or NV-NEW2) for the face of the identification target.

The face identification module 146 includes an N-dimensional vector (eg, NV-NEW1 or NV-NEW2) corresponding to the face of the identification target and a plurality of pre-stored N-dimensional vectors (NV-OG1 to NV-OGM) between each vector. The distance value can be calculated.

For example, the distance value between the vectors between the N-dimensional vector (eg, NV-NEW1) corresponding to the face image of the person to be identified and the plurality of pre-stored N-dimensional vectors (eg, NV-OG1) is as follows (Equation 1) can be calculated according to

(Formula 1)

Vd1=

In the same way, the face identification module 146 is a N-dimensional vector (eg, NV-NEW1) corresponding to the face image of the new registration target and the rest of the pre-stored N-dimensional vectors (NV-OG2 ~ NV-OGM) each of The distance between vectors can be calculated.

According to an embodiment, as the distance values between vectors are closer, the N-dimensional vectors illustrated in FIG. 2 may be located more closely.

The face identification module 146 identifies the person to be identified as a person corresponding to an N-dimensional vector whose distance value between vectors calculated from among a plurality of pre-stored N-dimensional vectors (NV-OG1 to NV-OGM) is within a reference distance value. can

According to an embodiment, the reference distance value may be a Euclidian distance, a cosine distance, or a Mahalanobis distance.

The parameter adjustment module 148 may adjust various parameters used by the face detection module 144 and the face identification module 146 to process an image frame.

According to an embodiment, the parameter adjustment module 148 may adjust a preset identification period. For example, the parameter adjustment module 148 may adjust a preset identification period when the face of the identification target is detected but not identified in at least one of the plurality of image frames. In this case, the parameter adjustment module 148 may reduce the preset identification period.

According to an embodiment, when adjusting the identification period, the parameter adjustment module 148 may adjust the first time limit for maintaining the adjusted identification period.

According to an embodiment, the parameter adjustment module 148 may adjust the second time limit for reducing the reference distance value or stopping the collection of image frames when the face of the subject to be identified is not identified for the first time limit.

A specific embodiment in which the first time limit and the second time limit are used will be described later with reference to FIGS. 4 and 5 .

4 is a flowchart of a face identification method according to an embodiment of the present invention. 5 is a diagram illustrating a process of performing face identification according to the face identification method of FIG. 4 .

Referring to FIGS. 4 and 5 , the face identification apparatus 100 may detect a face of an identification target included in an image frame at every preset identification cycle ( S410 ).

According to an embodiment, the face identification apparatus 100 may detect the face of the identification target included in at least one image frame among a plurality of image frames for each preset identification period TI1.

Each of time t1 to time t20 in FIG. 5 may mean a time at which an image frame is acquired. In this case, the preset identification period TI may be set to detect (or identify) the face of the identification target while skipping some image frames instead of all image frames. For example, in the case of FIG. 5, if the detection (or identification) of the face is performed at time t1, the detection (or identification) of the face with respect to the image frame acquired at time t2 and time t3 respectively is skipped, and the face at time t4 is skipped. Detection (or identification) may be set to be performed.

The face identification apparatus 100 may perform face identification on the image frame in which the face of the identification target is detected ( S420 ).

The face identification apparatus 100 may adjust the identification cycle when the face of the subject to be identified is detected in step S410 but not identified in step S420 ( S430 ).

According to an embodiment, the face identification apparatus 100 may maintain the identification cycle when the face of the identification target is not detected in step S410 .

According to an embodiment, when the face of the subject to be identified is detected in step S410 and identified in step S420, the face identification apparatus 100 may maintain the identification cycle.

Referring to FIG. 5 , it is assumed that the face is not detected in the first detection DT1 and the second detection DT2 , and the face of the identification target is detected but not identified in the third detection DT3 .

In this case, the face identification apparatus 100 may reduce the preset identification period TI1 to apply the changed identification period TI2 thereafter.

According to an embodiment, the face identification apparatus 100 may maintain the changed identification period TI2 until the first of a time when a face of an identification target is identified and a time when a preset first time limit TLMT1 has elapsed. .

In the case of FIG. 5 , it is assumed that the face of the identification target is identified in the sixth detection DT6 before the first time limit TLMT1 has elapsed. Thereafter, the face identification apparatus 100 may restore the changed identification period TI2 to the original identification period TI1, and accordingly, the seventh detection DT7 is to be performed according to the original identification period TI1. can

According to an embodiment, the face identification apparatus 100 may adjust the first time limit TLMT1 based on the distance between the N-dimensional vector corresponding to the identification target and the nearest N-dimensional vector among a plurality of pre-stored N-dimensional vectors. there is.

For example, the face identification apparatus 100 may adjust the first time limit to be longer as the distance between the nearest N-dimensional vectors is relatively close.

2 , in the case of an N-dimensional vector (NV-NEW1) corresponding to the first identification target, the distance between the nearest N-dimensional vectors has a value d1, and an N-dimensional vector (NV-NEW2) corresponding to the second identification target ), the distance between the nearest N Chauner vectors may have a value of d2 that is relatively farther than d1. In this case, the face identification apparatus 100 may adjust the first time limit to be longer when the first identification target is detected than when the second identification target is detected. For example, as the distance between the nearest N-dimensional vectors is relatively close, it may be within an error range due to motion blur, etc., so that stable face identification may be performed by increasing the first time limit.

According to an embodiment, in FIG. 5 , the face of the identification target is detected in the detections performed after the third detection DT6 (eg, DT4, DT5, DT6,??), but the first time limit TLMT1 is It is assumed that the face of the identification target is not identified until the lapse of time. In this case, the face identification apparatus 100 reduces the reference distance value used for face identification of the face identification module 146 for the second time limit TLMT2, or stops the collection of image frames for the second time limit TLMT2. can be stopped For example, if the face of the subject to be identified is not identified until the first time limit TLMT1 has elapsed, it is determined that it is highly likely that the subject to be identified is repeatedly requesting identification, and the second time limit ( During TLMT2), the identification procedure can be stopped by reducing the reference distance value used for face identification to make face identification a strict criterion or by stopping the collection of image frames during the second time limit (TLMT2).

The face identification method according to an embodiment of the present invention may be implemented as a program code and stored in a memory (eg, 120 in FIG. 1 ), and the memory (eg, 120 in FIG. 1 ) is provided by a processor (eg, 140 in FIG. 1 ). In combination with , it is possible to perform the face identification method according to an embodiment of the present invention.

As mentioned above, although the present invention has been described in detail with reference to a preferred embodiment, the present invention is not limited to the above embodiment, and various modifications and changes can be made by those skilled in the art within the technical spirit and scope of the present invention. This is possible.

100: face identification device
110: image capture unit
120: memory
130: communication interface
140: processor

Claims (10)

detecting a face of an identification target included in at least one image frame among the plurality of image frames for each of the plurality of collected image frames;
performing face identification with respect to an image frame in which the face of the subject to be identified is detected; and
according to the result of performing the face identification, adjusting the identification period when the face of the identification target is detected but not identified;
The step of detecting the face includes generating an N-dimensional vector (where N is a natural number greater than or equal to 2) corresponding to the face image from a face image including the face of the identified subject,
The step of adjusting the identification period may include reducing the identification period when the face of the identification target is detected but not identified according to a result of performing the face identification, and then, when the face of the identification target is detected, the changed identification A method of face identification, to which a cycle is applied.
According to claim 1,
The step of adjusting the identification period is,
When the face of the identification target is not detected in the at least one image frame, the identification period is maintained.
delete According to claim 1,
The step of adjusting the identification period is,
When the face is not identified according to the result of performing the face identification, the reduced identification period is maintained until the first of the face is identified and the time when a preset first time limit has elapsed. method.
5. The method of claim 4,
The step of performing the face identification comprises:
calculating a distance value between the generated N-dimensional vector and each of a plurality of previously stored N-dimensional vectors; and
and identifying the person to be identified as a person corresponding to an N-dimensional vector in which the calculated inter-vector distance value is within a reference distance value from among a plurality of pre-stored N-dimensional vectors.
6. The method of claim 5,
The face identification method comprises:
The method further comprising: adjusting the first time limit based on a distance between the N-dimensional vector corresponding to the identification target and the nearest N-dimensional vector among the plurality of pre-stored N-dimensional vectors.
7. The method of claim 6,
The step of adjusting the first time limit,
Adjusting the first time limit to be longer as the distance between the nearest N-dimensional vectors is relatively close, the face identification method.
6. The method of claim 5,
The face identification method comprises:
If the face of the identification target is not identified until the preset first time limit has elapsed, the reference distance value is reduced for a second time limit or the collection of image frames is stopped during the second time limit. identification method.
a face detection module configured to detect a face of a subject to be identified included in at least one image frame among the plurality of image frames for each of the plurality of collected image frames;
a face identification module for performing face identification with respect to an image frame in which a face of a subject to be identified is detected; and
a parameter adjustment module for adjusting the identification period when a face is not identified according to a result of performing the face identification;
The face detection module generates an N-dimensional vector (where N is a natural number greater than or equal to 2) corresponding to the face image from a face image including the face of the identified subject when detecting the face of the identified subject,
The parameter adjustment module reduces the identification period when the face of the identification subject is detected but not identified according to a result of performing the face identification, and then, when the face of the identification subject is detected, the changed identification period is applied Being a face identification device.
A program stored in a medium for performing a face identification method in combination with a processor,
detecting a face of an identification target included in at least one image frame among the plurality of image frames for each of the plurality of collected image frames;
performing face identification with respect to an image frame in which the face of the subject to be identified is detected; and
and program code for performing the step of adjusting the identification period when the face of the identification target is detected but not identified according to the result of performing the face identification,
The step of detecting the face includes generating an N-dimensional vector (where N is a natural number greater than or equal to 2) corresponding to the face image from a face image including the face of the identified subject,
The step of adjusting the identification period may include reducing the identification period when the face of the identification target is detected but not identified according to a result of performing the face identification, and then, when the face of the identification target is detected, the changed identification A program stored on a medium to which a cycle is applied

Images