KR20090119565A - Apparatus for face detect and recognition based on embedded system and face dectect and recognition in using the apparatus - Google Patents

Abstract

PURPOSE: An embedded system-based face search and recognition device and a method thereof are provided to recognize/search the face video information in real time by using the AdaBoost algorithm. CONSTITUTION: The feature point of the received video information is analyzed through the AdaBoost algorithm(S910). If the received video information is the face information, the searched face image information is processed as the specific size(S920). The face part of the processed face information is recognized through the PCA(Principal Component Analysis) algorithm(S930). The face video information is ported with the embedded type(S940). The ported face video information is transmitted to the embedded application(S950).

Description

Face detection and recognition device based on embedded system and face detection and recognition method using the device {APPARATUS FOR FACE DETECT AND RECOGNITION BASED ON EMBEDDED SYSTEM AND FACE DECTECT AND RECOGNITION IN USING THE APPARATUS}

The present invention relates to an apparatus for searching and recognizing a face based on an embedded system and a method for searching and recognizing a face using the apparatus, and more particularly, searching for face image information from image information using an Adaboost algorithm, and searching the face image. The present invention relates to a face searching and recognizing apparatus capable of recognizing a face region using a principal component analysis (PCA) algorithm from information and to porting the face image information available to an embedded system, and a face searching and recognizing method using the apparatus.

Today, as the digital technology develops, a technology for analyzing image information and dividing it into a specific area or a specific part is being developed.

Among the analysis techniques, the face recognition technique is a technology that is applied not only to a digital camera but also to a device that performs a security technique, and is being researched and developed in various ways in various ways.

In addition, in order to accurately recognize the face, a technique of accurately searching for face image information in image information should be accompanied. Conventional face searching techniques use skin color based algorithm and neural network algorithm.

However, the skin color algorithm and the neural network algorithm do not search face image information accurately, or the search time is too long.

Accordingly, there is a need for a face searching and recognizing apparatus that can search for or recognize face image information in real time by analyzing image information at a higher speed so as to be matched with a digital device.

The present invention has been made to solve the above-mentioned problems of the prior art, and an object thereof is to provide a method for searching and recognizing face image information based on an embedded system.

Another object of the present invention is to provide a method for searching face image information in real time using the AdaBoost algorithm.

In addition, an object of the present invention is to provide a method for recognizing accurate face image information by using a principal component analysis (PCA) algorithm.

In order to achieve the above object and solve the above-mentioned problems of the prior art, the embedded system-based face searching and recognition method according to an embodiment of the present invention, the feature point of the image information received through the AdaBoost algorithm Analyzing whether the image information is face image information; if the received image information is face image information, processing the searched face image information to a specific size according to a preset setting; Recognizing a face region through a Principal Component Analysis (PCA) algorithm of the image information and porting the face image information recognized by the face region in an embedded format.

The searching of the face image information according to the present invention may include: a first step of initializing a weight value of the received image information, a second step of normalizing the initialized weight value, and the specific feature point A third step of calculating a resultant value and an error value of a classifier for the second step; a fourth step of selecting a classifier having the smallest error value among the specific classifiers and storing the result value in a database; When storing, repeating the fifth step of updating the weight value, the first step to the fifth step for a predetermined number of times, and combining the result values of the classifier stored in the fourth step into a result value of the final classifier. And a seventh step of determining whether the face image information is based on the result of the final classifier.

The recognizing of the face region may include obtaining a learning face image information result value obtained by converting image information recognized as one or more face image information into a column vector format, and averaging the image information recognized as the face image information. Obtaining a face image information result value, storing a face image information comparison value, which is a difference between the average face image information result value, in a database, and sorting the face image information result value by a covariance matrix of the face image information comparison value Calculating a face image information matrix value, calculating an eigenvalue of the face image information matrix value and a magnitude of an eigenvector, and arranging the eigenvectors in size order based on the eigenvalues.

The recognizing a face region through a principal component analysis (PCA) algorithm of the processed face image information of the present invention may include: projecting the learning face image information result value into the eigenvector; Projecting the resulting face image information result into the eigenvector, calculating a Euclidean distance between the projected learning face image information and the projected processed face image information, and finally recognizing that the calculated Euclidean distance is minimum The method may further include determining the face image information.

In order to achieve the above object and solve the above-mentioned problems of the prior art, the embedded system-based face search and recognition device according to an embodiment of the present invention, the feature point of the image information received through the AdaBoost algorithm A face searching unit for searching whether the face image information is analyzed by analyzing the image processing unit, and if the received image information is face image information as a result of the search, an image processing unit processing the found face image information according to a preset size to a specific size; A face recognition unit for recognizing a face region through a principal component analysis (PCA) algorithm among the processed face image information, and a porting controller for porting the face image information in which the face region is recognized in an embedded format, and the ported face. And a face image information transmitter for transmitting the image information to the embedded application means.

According to the present invention, the facial image information based on the embedded system can be searched and recognized.

According to the present invention, the face information can be searched for in real time using the AdaBoost algorithm.

In addition, according to the present invention, it is possible to recognize the correct face image information by using the principal component analysis (PCA) algorithm.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings and the contents described in the accompanying drawings, but the present invention is not limited or limited to the embodiments.

On the other hand, in describing the present invention, when it is determined that the detailed description of the related known function or configuration may unnecessarily obscure the subject matter of the present invention, the detailed description thereof will be omitted. Terminology used herein is a term used to properly express a preferred embodiment of the present invention, which may vary according to a user, an operator's intention, or a custom in the field to which the present invention belongs. Therefore, the definitions of the terms should be made based on the contents throughout the specification.

1 is a block diagram illustrating the overall configuration of an apparatus for face searching and recognizing based on an embedded system according to an exemplary embodiment of the present invention.

The present invention provides a real-time face search and recognition method based on the embedded system. That is, based on the image information received by the embedded system, the face search and recognition apparatus 100 searches for face image information using the AdaBoost algorithm, recognizes the face image information using the principal component analysis algorithm, and then uses the embedded application means. Provided is a method of porting and providing the face image information in an embedded system format reproducible at 200.

In this case, the face searching and recognizing apparatus 100 includes various means for processing image information such as a camera applied to an embedded system base.

As described above, the facial image information searching and recognizing apparatus 100 according to the present invention is described in detail according to the sequential flow of image information until the image information is collected and recognized as the facial image information with reference to FIGS. 2 to 8. Let's explain.

2 is a block diagram illustrating a configuration of an apparatus for face searching and recognizing based on an embedded system according to an exemplary embodiment of the present invention.

First, the face searching unit 110 analyzes a feature of the image information received through the AdaBoost algorithm and searches whether the face image information is face image information.

That is, the face searching unit 110 analyzes the image information received by the face searching and recognizing apparatus 100 by capturing an arbitrary image by a user using a photographing means connected to the face searching and recognizing apparatus 100. Search for face image information.

At this time, in the present invention, by analyzing the image information through the AdaBoost algorithm, it is possible to search whether the face image information in real time.

In the present invention, the face image information search function searches for the face image information by analyzing a rectangular feature value of the received image information through the AdaBoost algorithm.

In the present invention, the rectangular feature point extraction is based on the Haar basis function to quickly analyze the facial features.

3 is a diagram illustrating a feature point analysis type according to an embodiment of the present invention.

That is, in the face image information search method of the present invention, as shown in FIG. do. For example, when the size of the window to be searched is 24x24, the number of feature points that can be obtained by changing the size and position of the rectangular feature points is about 40,000.

A process of obtaining the rectangular feature point value finally obtained from the face image information search method provided by the present invention through the AdaBoost algorithm will now be described with reference to FIG. 4.

4 is a diagram illustrating an example of searching for face image information using the AdaBoost algorithm according to an embodiment of the present invention.

First, a weight value of the received image information is initialized (first step).

Next, the initialized weight value is normalized (second step).

Next, a result value and an error value of a classifier for the feature point are calculated (third step).

Next, the classifier having the smallest error value is selected among the classifiers, and the result value is stored in the database 131 (step 4).

Next, when the result value of the selected classifier is stored, the weight value is updated (step 5).

Next, the first to fifth steps are repeated for a predetermined number of times, and the resultant values of the classifier stored in the fourth step are combined and processed into the final value of the final classifier (sixth step).

That is, according to the face image information searching method of the present invention, a repetitive block is generated as shown in FIG. 4 according to the result value of the final classifier through the sixth step, and analyzed to determine whether it is the face image information. Can be determined in real time (step 7).

In this case, the result value h (x) of the final classifier may be calculated through Equation 1 below.

는 상기 무게값이다. In this case, h (1) is a final classifier result value in which the image information is determined to be face image information, and h (0) is a final classifier value in which it is determined that the image information is not face image information.

Is the weight value.

Next, when the received image information is face image information, the image processing unit 120 processes the found face image information to a specific size according to a preset setting.

The image processing unit 120 of the present invention receives the size setting information of the searched face image information, converts the searched face image information to an appropriate size according to the received size setting information, and then converts the converted face. By processing the image information to a predetermined standard size, it is possible to provide a format that can accurately recognize the searched face image information.

Next, the face recognition unit 130 recognizes a face region of the processed face image information through a Principal Component Analysis (PCA) algorithm.

According to the present invention, face image information which is the basis for face image information recognition can be learned and stored in advance, and final face image information can be recognized based on the learned face image information using the principal component analysis algorithm. have.

Looking at the process of learning the face image information of the present invention through the principal component analysis algorithm as follows.

First, a result of learning face image information obtained by converting image information recognized as one or more face image information into a column vector format is obtained.

Next, an average face image information result value of the image information recognized as the face image information is obtained.

)을 데이터베이스(131)에 저장한다.Next, a face image information comparison value that is a difference between the learning face image information result value and the average face image information result value (

) Is stored in the database 131.

)의 공분산 행렬로 정렬한 얼굴 영상 정보 행렬값(Ω)을 계산한다.Next, the face image information comparison value (

The face image information matrix value (Ω) arranged by the covariance matrix of

Next, the eigen value of the face image information matrix value? And the magnitude of the eigenvector are calculated.

Finally, the eigenvectors may be arranged as shown in FIG. 5 in order of magnitude based on the eigenvalues.

5 is a diagram illustrating an example of classifying result values of image information using a principal component analysis algorithm according to the present invention.

According to the present invention, a face region may be recognized through a principal component analysis (PCA) algorithm among the processed face image information based on the learned face image information.

FIG. 6 is a diagram illustrating an example of unique face image information extracted through a principal component analysis algorithm according to an embodiment of the present invention.

For example, in the present invention, since the learned eigenvector is similar to the face shape as shown in FIG. 6, it can be referred to as eigenface image information, and the processed eigenvector is processed based on the result value of the learned image information. By comparing the face image information, accurate face image information can be recognized as follows.

First, the learning face image information result value is projected on the eigenvector, and the processed face image information result value is projected on the eigenvector.

In this case, the Euclidean distance between the projected learning face image information and the projected processed face image information may be calculated, and it may be determined that the calculated Euclidean distance is minimum as the final recognized face image information.

7 illustrates an example of final face image information recognized through a principal component analysis algorithm according to an embodiment of the present invention.

In addition, as shown in FIG. 7, the final recognized face image information recognized through the principal component analysis algorithm provides robustness to illumination.

Next, the porting controller 140 ports the face image information in which the face region is recognized in an embedded format.

The present invention provides a method for porting the recognized face image information into a reproducible format in the embedded application means 200 by searching and recognizing face image information through the AdaBoost algorithm and the principal component analysis algorithm as described above.

8 is a flowchart illustrating a method for porting to an embedded form according to an embodiment of the present invention.

That is, the present invention portes the recognized face image information in a format reproducible by the embedded application means 200 as shown in FIG. 8. An embodiment of the porting process will be described below.

First, Wince (Platform Builder) is installed on the PC, and the target board is configured on the Platform Builder (PB) of the PC to add components, create executable binary files (NK.bin, NK.nb0), and configure necessary expansion boards. .

Next, create and install the target board's SDK from P.B on the PC, and install the development tool (EVC4.0).

Next, set communication and boot-loader settings between the PC and the target board.

Next, Boot-Loader communication between PC and target board is composed of hyper terminal, and executable binary file created on OS, that is, P.B, is installed on target board.

Finally, the porting process is completed by creating and testing the application on the target board. At this time, the application of the program written in MFC can be converted into the application to be converted to the target board by EVC, downloaded, and tested.

The face image information of which porting is completed in the above process is transmitted to the embedded application means 200 through the face image information transmitter 150.

As described above, the present invention provides a method for searching and recognizing face image information through AdaBoost algorithm and principal component analysis algorithm, and providing the face image information as information suitable for the format of the embedded application means 200. This will be described sequentially with reference to FIGS. 9 to 13.

In this case, since the method corresponds to the face searching and recognizing apparatus 200, since the method includes all the functional elements of the apparatus, detailed description thereof will be omitted.

9 is a flowchart illustrating a face searching and recognition method based on an embedded system according to an exemplary embodiment of the present invention.

First, the face search unit 110 of the present invention analyzes a feature of the image information received through the AdaBoost algorithm as shown in FIG. 9 and searches whether the face image information is face image information (S910).

In operation S910, as shown in FIG. 10, face image information may be searched from the image information by the following process.

10 is a flowchart illustrating a method of searching for face image information according to an embodiment of the present invention.

First, the face search unit 110 initializes a weight value of the received image information (S1010).

Next, the face search unit 110 normalizes the initialized weight value (S1020).

Next, the face searching unit 110 calculates a result value and an error value of a classifier for the feature point (S1030).

Next, the face searching unit 110 selects the classifier having the smallest error value among the classifiers and stores the result value in the database 131 (S1040).

Next, when the face search unit 110 stores the result value of the selected classifier, the face search unit 110 updates the weight value (S1050).

Next, the face searching unit 110 repeats the first step S1010 to the fifth step S1050 for a predetermined number of times, and combines the result values of the classifier stored in the fourth step S1040 to determine the final classifier. It is processed to the result of (S1060).

Finally, the face searching unit 110 determines whether the face image information is the result of the final classifier (S1070).

Next, when the received image information is face image information, the image processing unit 120 processes the searched face image information to a specific size according to a preset setting (S920).

In operation S920, as illustrated in FIG. 11, the detected face image information may be processed to a specific size.

11 is a flowchart illustrating a method of processing the found face image information to a specific size according to an embodiment of the present invention.

First, the image processor 120 receives size setting information of the searched face image information (S1110).

Next, the image processing unit 120 converts the searched face image information to an appropriate size according to the received size setting information (S1120).

Finally, the image processing unit 120 processes the converted face image information to a predetermined standard size (S1130).

Next, the face recognition unit 130 recognizes the face region through the principal component analysis (PCA) algorithm among the processed face image information (S130).

In operation S930, as illustrated in FIG. 12, a face region may be recognized among face image information processed by the following process.

12 is a flowchart illustrating a learning method of recognizing a face region of the processed face image information according to an embodiment of the present invention.

First, the face recognition unit 130 obtains a learning face image information result value obtained by converting image information recognized as one or more face image information into a column vector format (S1210).

Next, the face recognizing unit 130 obtains an average face image information result value of the image information recognized as the face image information (S1220).

Next, the face recognizing unit 130 stores the face image information comparison value, which is a difference between the average face image information result value, in the database 131 in the learning face image information result value (S1230).

Next, the face recognition unit 130 calculates face image information matrix values arranged in the covariance matrix of the face image information comparison values (S1240).

Next, the face recognition unit 130 calculates the eigenvalue of the face image information matrix value and the magnitude of the eigenvector (S1250).

Next, the face recognition unit 130 sorts the eigenvectors in size order based on the eigenvalues (S1260).

According to the present invention, after aligning the eigenvectors of the face image information for the initial learning of the face image information in the above manner, and projecting the image information to be recognized as the face image information, the face of the face image information is finally as follows. Area can be recognized.

FIG. 13 is a flowchart illustrating a method of recognizing a face region of processed face image information according to an exemplary embodiment of the present invention.

First, the face recognition unit 130 projects the learning face image information result value into the eigenvector (S1310).

Next, the face recognition unit 130 projects the processed face image information result value into the eigenvector.

Next, the face recognition unit 130 calculates a Euclidean distance between the projected learning face image information and the projected processed face image information (S1320).

Finally, the face recognition unit 130 determines that the calculated Euclidean distance is the minimum as final recognized face image information (S1330).

Next, the porting controller 140 ports the face image information in which the face region is recognized in an embedded format.

Finally, the face image information transmitter 150 transmits the ported face image information to the embedded application means 200.

Embodiments according to the present invention can be implemented in the form of program instructions that can be executed by various computer means can be recorded on a computer readable medium. The computer readable medium may include program instructions, data files, data structures, etc. alone or in combination. Program instructions recorded on the media may be those specially designed and constructed for the purposes of the present invention, or they may be of the kind well-known and available to those having skill in the computer software arts. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tape, optical media such as CD-ROMs, DVDs, and magnetic disks, such as floppy disks. Magneto-optical media, and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like. Examples of program instructions include not only machine code generated by a compiler, but also high-level language code that can be executed by a computer using an interpreter or the like. The hardware device described above may be configured to operate as one or more software modules to perform the operations of the present invention, and vice versa.

As described above, the present invention has been described by way of limited embodiments and drawings, but the present invention is not limited to the above embodiments, and those skilled in the art to which the present invention pertains various modifications and variations from such descriptions. This is possible. Therefore, the scope of the present invention should not be limited to the described embodiments, but should be determined not only by the claims below but also by the equivalents of the claims.

1 is a block diagram illustrating the overall configuration of an apparatus for face searching and recognizing based on an embedded system according to an exemplary embodiment of the present invention.

2 is a block diagram illustrating a configuration of an apparatus for face searching and recognizing based on an embedded system according to an exemplary embodiment of the present invention.

3 is a diagram illustrating a feature point analysis type according to an embodiment of the present invention.

4 is a diagram illustrating an example of searching for face image information using the AdaBoost algorithm according to an embodiment of the present invention.

5 is a diagram illustrating an example of classifying result values of image information using a principal component analysis algorithm according to the present invention.

FIG. 6 is a diagram illustrating an example of unique face image information extracted through a principal component analysis algorithm according to an embodiment of the present invention.

7 illustrates an example of final face image information recognized through a principal component analysis algorithm according to an embodiment of the present invention.

8 is a flowchart illustrating a method for porting to an embedded form according to an embodiment of the present invention.

9 is a flowchart illustrating a face searching and recognition method based on an embedded system according to an exemplary embodiment of the present invention.

10 is a flowchart illustrating a method of searching for face image information according to an embodiment of the present invention.

11 is a flowchart illustrating a method of processing the found face image information to a specific size according to an embodiment of the present invention.

12 is a flowchart illustrating a learning method of recognizing a face region of the processed face image information according to an embodiment of the present invention.

FIG. 13 is a flowchart illustrating a method of recognizing a face region of processed face image information according to an exemplary embodiment of the present invention.

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

110: face search unit

120: image processing unit

130: face recognition unit

131: database

140: porting control unit

150: face image information transmission unit

Claims (10)

Analyzing whether a feature of the image information received through the AdaBoost algorithm is face image information; If the received image information is face image information, processing the searched face image information to a specific size according to a preset result; Recognizing a face region through a principal component analysis (PCA) algorithm among the processed face image information; And Porting the face image information recognized by the face region in an embedded format Face detection and recognition method based on the embedded system comprising a. The method of claim 1, Transmitting the ported face image information to an embedded application means Face detection and recognition method based on the embedded system, characterized in that it further comprises. The method of claim 1, The searching for the face image information may include: An embedded system-based face searching and recognition method according to claim 1, wherein the rectangular feature point of the received image information is analyzed through the AdaBoost algorithm to search whether the image information is face image information. The method of claim 1, The searching for the face image information may include: A first step of initializing a weight value for the received image information; A second step of normalizing the initialized weight value; Calculating a resultant value and an error value of a classifier for the feature point; A fourth step of selecting a classifier having the smallest error value among the classifiers and storing a result value in a database; A fifth step of updating the weight value when the result value of the selected classifier is stored; A sixth step of repeating the first to fifth steps for a predetermined number of times and combining the result values of the classifier stored in the fourth step into a result value of the final classifier; And A seventh step of determining whether the face image information is based on a result value of the final classifier; Face detection and recognition method based on the embedded system comprising a. The method of claim 4, wherein The result of the final classifier is an embedded system-based face detection and recognition method, characterized in that the value (h (x)) calculated through the following equation (2).

(H (1) is a final classifier result value in which the image information is determined to be face image information, and h (0) is a final classifier value in which it is determined that the image information is not face image information,

Is the weight value.) The method of claim 1, The processing of the searched face image information to a specific size according to a preset may include: Receiving size setting information of the searched face image information; Converting the found face image information into an appropriate size according to the received size setting information; And Processing the converted face image information to a preset standard size Face detection and recognition method based on the embedded system comprising a. The method of claim 1, Recognizing a face region through a principal component analysis (PCA) algorithm of the processed face image information may include: Obtaining a learning face image information result value obtained by converting image information recognized as at least one face image information into a column vector format; Obtaining an average face image information result value of the image information recognized as the face image information; Storing a face image information comparison value, which is a difference between the average face image information result value, and a database in the learning face image information result value; Calculating face image information matrix values arranged in a covariance matrix of the face image information comparison values; Calculating eigenvalues of the face image information matrix values and magnitudes of eigenvectors; And Sorting the eigenvectors in magnitude order based on the eigenvalues Face detection and recognition method based on the embedded system comprising a. The method of claim 7, wherein Recognizing a face region through a principal component analysis (PCA) algorithm of the processed face image information may include: Projecting the learning face image information result value into the eigenvector; Projecting the processed facial image information result into the eigenvector; Calculating a Euclidean distance between the projected learning face image information and the projected processed face image information; And Determining that the calculated Euclidean distance is the minimum as final recognized face image information Face detection and recognition method based on the embedded system, characterized in that it further comprises. A computer-readable recording medium in which a program for executing the method of any one of claims 1 to 8 is recorded. A face searching unit that analyzes a feature of the image information received through the AdaBoost algorithm and searches whether the face image information is face image information; An image processing unit for processing the searched face image information to a specific size according to a preset setting when the received image information is face image information as a result of the searching; A face recognition unit recognizing a face region through a principal component analysis (PCA) algorithm among the processed face image information; A porting controller for porting the face image information in which the face region is recognized in an embedded format; And Face image information transmitter for transmitting the ported face image information to an embedded application means Face detection and recognition device based on the embedded system comprising a.

Images