KR101150944B1 - Apparatus and Method of searching for target based on matching possibility - Google Patents

Abstract

A target search apparatus based on matching possibility, a method thereof, and a recording medium therefor are disclosed. A target searching apparatus based on matching possibility according to an embodiment of the present invention includes a target information database storing a face shape and a body shape of a target desired by a first user; Input means for receiving priority information between a face shape and a body shape of the target; A camera for receiving an image of a second user; A preprocessor for preprocessing an image of the second user to recognize only a facial part and then recognizing a facial shape and distinguishing an area corresponding to an arm, a leg, and a body of the second user from the image of the second user; Calculating a correlation coefficient by comparing with a target face shape stored in the target information database, measuring a head size in the image of the second user, calculating a ratio of each part of the body using the measured head size, A similarity determination unit for comparing the body shape of the target body shape with the target body shape stored in the target information database to calculate a first degree of similarity; And a target matching determination unit that calculates a target matching index by applying a weight to the correlation coefficient and the first similarity according to the priority information.

Description

Field of the Invention [0001] The present invention relates to a target searching apparatus based on matching possibility, a method thereof, and a recording medium thereof.

More particularly, the present invention relates to a target search apparatus based on matching possibility, a method thereof, and a recording medium therefor.

As the importance of recognition through video is increasing day by day, technologies using human face are increasing. In particular, face detection and face recognition are used in various fields such as biometrics, security systems, and video search. The most active area for researching face detection function is digital camera. In the case of a camera, the face detection function is becoming a basic function in accordance with the characteristic that a person is a main object to be photographed. In addition, image retrieval service using the face similarity of the person is being strengthened.

Mobile phones and smart phones basically have built-in digital cameras. Facial search service, which was previously provided on the web, is also being implemented in the mobile environment. However, the conventional face search service only searches for similar photographs based on several feature points, so accuracy is low and its utilization is limited.

A first object of the present invention is to provide a target searching apparatus based on matching possibility that can inform a user of various characteristics of a person whether a photographed object is a person desired by a user based on accurate and fast image recognition I have to.

A second object of the present invention is to provide a target searching method based on the matching possibility applied to a target searching apparatus based on the matching possibility.

A third object of the present invention is to provide a recording medium readable by a computer system for recording a program for executing a target searching method based on the matching possibility in a computer system.

According to an aspect of the present invention, there is provided a target search apparatus based on matching possibility according to an embodiment of the present invention includes a target information database storing a face shape and a body shape of a target desired by a first user; Input means for receiving priority information between a face shape and a body shape of the target; A camera for receiving an image of a second user; A preprocessor for preprocessing an image of the second user to recognize only a facial part and then recognizing a facial shape and distinguishing an area corresponding to an arm, a leg, and a body of the second user from the image of the second user; Calculating a correlation coefficient by comparing with a target face shape stored in the target information database, measuring a head size in the image of the second user, calculating a ratio of each part of the body using the measured head size, A similarity determination unit for comparing the body shape of the target body shape with the target body shape stored in the target information database to calculate a first degree of similarity; And a target matching determination unit that calculates a target matching index by applying a weight to the correlation coefficient and the first similarity according to the priority information.

In order to achieve the second object, in a target search method based on matching possibility according to an embodiment of the present invention, in a terminal including a target information database storing a face shape and a body shape of a target desired by a first user, Receiving priority information of a face shape and a body shape of the body; Receiving an image of a second user; Calculating a correlation coefficient by comparing a face shape of the second user with a target face shape stored in a target information database; A head size is measured in the image of the second user, a ratio of each part of the body is calculated using the measured head size, a body type is classified according to the calculated ratio, and the body shape is compared with the target body type stored in the target information database. 1 < / RTI > And calculating a target matching index by applying a weight to the correlation coefficient and the first similarity according to the priority information.

According to the embodiments of the present invention, since the photographed object can be informed to the user by reflecting various characteristics of the person whether the photographed object is the person desired by the user or not based on accurate and fast image recognition, And the like.

FIG. 1A is a block diagram of a target searching apparatus based on matching possibility according to an embodiment of the present invention.
FIGS. 1B to 1D show an example of a process of recognizing a face shape in FIG. 1A.
2 is a block diagram of a target searching apparatus based on matching possibility according to another embodiment of the present invention.
3 is a flow chart of a target searching method based on matching possibility according to an embodiment of the present invention.
4 is a flowchart of a target search method based on matching possibility according to another embodiment of the present invention.
FIG. 5 shows an example of a process of configuring a database in FIG.

Hereinafter, preferred embodiments of the present invention will be described with reference to the drawings. However, the following embodiments of the present invention may be modified into various other forms, and the scope of the present invention is not limited to the embodiments described below.

FIG. 1A is a block diagram of a target searching apparatus based on matching possibility according to an embodiment of the present invention. The target searching apparatus of FIG. 1A is assumed to be embedded in a user terminal such as a mobile phone, a smart phone, a PDA, a notebook, a laptop, and a dedicated terminal manufactured to determine whether a target is matched.

The camera 110 is a digital camera that captures an object and generates or receives an image of a second user.

The preprocessor 120 preprocesses the image of the second user and recognizes the facial shape after detecting only the face portion. FIG. 1B shows an example of detecting a face part through preprocessing. The Haar-like Feature algorithm or the AdaBoost algorithm can be used to detect only part of the face in the image, to recognize the face shape or to compare with other people's face. The Haar-like Feature algorithm is an algorithm that compresses or simplifies an image to compare it with a previously selected target image. &Quot; A " and " B " in Fig. 1C represent the edge characteristic, " C " represents the line characteristic, and " D " represents the Haar-

The preprocessor 120 distinguishes an area corresponding to an arm, a leg, and a body of a second user from an image of a second user, and a Haar-like feature algorithm or AdaBoost algorithm can be used for this part.

The input means 130 receives the priority information of the face shape and the body shape of the target, and may have various forms such as a keypad, a keyboard, a touch panel, and other pointing devices.

The target matching determination unit 140 compares the target face shape stored in the target information database 150 and calculates a correlation coefficient. If you compare two images one by one in pixel units, it will take a long time. Therefore, Haar-like Features defines a number of boxes of a certain size as shown in FIG. 1C, and the number of points included in each box is equal to or smaller than the threshold value within the range of the threshold value do. The comparison results are divided into three categories. There are few cases, the same case, and many cases. According to the result, the corresponding position of each box is painted with a specific color to make the image simple. Specifically, for example, if a box is defined as " A " in Fig. 1C, the number of points in each box in two images is compared. FIG. 1D shows an example applied to actual face recognition. The size and position of the box are determined by learning from a plurality of face data. The result of this algorithm is a correlation coefficient indicating how much the two images differ in each box area. The offset factor computes a comparable degree of similarity between the two images. As a result, the correlation coefficients between images compressed with Haar-like features are compared to show how much difference occurs with the other images. In detail, the degree of resemblance to each part of a person's eyes, nose, mouth, chin, body, and style can be compared and quantified in both images.

The target matching determining unit 140 measures the head size in the image of the second user, calculates the ratio of each part of the body using the measured head size, and classifies the body according to the calculated ratio. For example, this process can be implemented through the following process. Using the whole body image from head to foot, Haar-like Features are recognized and extracted from the recognized body only. After calculating the size from the extracted hair, the body lengths of the tall, short, seventh, eighth, dull, plump, and glamor types can be obtained by comparing the remaining lengths. The criteria for distinguishing between body information and body shape can be predefined in the form of a table by standardizing the body length or the ratio.

The target matching determination unit 140 calculates a first similarity degree that indicates a degree of similarity compared with the target object type stored in the target information database 150. [ Hereinafter, the first similarity degree, the second similarity degree, the third similarity degree, and the like are obtained by comparing the characteristics obtained by the target matching determination unit 140 and the characteristics stored in the target information database 150, the ratio of the number of identical / similar elements among the total number of elements, or can be obtained by using a probability function used in statistical analysis.

The target matching determining unit 140 calculates the target matching index by applying the weight to the correlation coefficient, the first similarity, the second similarity, and the third similarity according to the received priority information. For example, if the priority information indicates that the body type takes precedence over the face type, a weight of 0.6 may be assigned to the body type and a weight of 0.4 to the face type.

The target information database 150 stores the face shape, the body shape, and the like of the target desired by the first user. The method of generating / constructing the information about the face shape, the body shape, and the like of the target is the same as the method of extracting the information about the face shape, the body shape, etc. of the second user.

In another embodiment of the present invention, a user selects and combines partial images such as a head image, a snow image, a nose image, a mouth image, a face image, and the like stored in advance in the target information database 150 through the input means 130 A method of determining the face shape, the body shape, etc. of the target can be used.

2 is a block diagram of a target searching apparatus based on matching possibility according to another embodiment of the present invention.

The sensor unit 160 includes a gyro sensor, an acceleration sensor, an infrared sensor, an ultrasonic sensor, and the like. Since the whole body image for determining the body shape may have an error depending on the photographing angle, the preprocessor 120 can correct such an error using a gyro sensor and an acceleration sensor. The error of the distance can be corrected using the measurement value of the infrared sensor.

The matching test unit 170 performs a questionnaire type matching test on the second user. Hereinafter, an example of the matching test method will be described. Matching tests use surveys or statistical data to distinguish between types of people through entertainment orientation, personality, psychology, blood type tests, and so on. The type of problem may vary depending on the program you are implementing. For example, as shown in Table 1, MBTI psychological tests can be used to divide the personality into one of 16 categories.

Sensual Intuitive Introverted
ISTJ ISFJ INFJ INTJ Judgmental ISTP ISFP INFP INTP Cognitive Extrovert
ESTP ESFP ENFP ENTP Cognitive ESTJ ESFJ ENFJ ENTJ Judgmental Accidental Emotional Emotional Accidental

In addition, as shown in Table 2, statistical analysis of blood type compatibility information and the like can be reflected.

Woman Man A O B AB A 70% 95% 20% 65% O 90% 40% 80% 30% B 25% 75% 65% 80% AB 50% 35% 85% 90%

On the other hand, a result of inputting "O" or "X" for each item by the first user is compared with the input result of the second user while performing an entertaining psychological test or personality test showing items such as Tables 3 and 4 The matching test may be performed.

Item survey One I like to go straight home after work. 2 I go to work before going to work after work. 3 It is not lonely even if you eat rice alone. 4 I can not remember where I left my favorite card.

Item survey One Emotions are revealed on the face. 2 I like new attempts. 3 I tend to do things at that time rather than plan. 4 I like meeting new people.

The input means 130 receives the user information including the response value of the questionnaire, personal information, and the like from the second user during the matching test at the terminal.

The preprocessor 120 not only distinguishes the areas corresponding to the arms, legs, and bodies of the second user from the images of the second user but also distinguishes the hair of the second user from the images of the second user, Style information can be detected. Haar-like feature algorithm or AdaBoost algorithm can also be used in this part. Hereinafter, an exemplary method of distinguishing styles will be described. The fashion style can be distinguished by hair, clothes and other accessories. Haar-like features are used to detect the accessibility of hair, clothing, bags, etc., and the correlation coefficient is used to analyze the types of hairstyle, accessories, and costumes, and classified into suits, casuals, and hip- Style can be inferred. In addition, it is also possible to determine whether a proper style combination is obtained by using hair style, costume, and accessory information, and may affect the determination of matching with a target according to a user characteristic. In addition to image recognition, information about the style may be input through the response of the question item.

The target matching determining unit 140 compares the style stored in the target information database 150 and calculates the second similarity.

The target matching determination unit 140 determines the type of the second user according to the result of the matching test and compares the type of the second user with the target type stored in the target information database 150 to calculate the third similarity. The type of the first user and the type of the second user are compared. Based on the above data, the higher the similarity of a specific item, the higher the degree of similarity of a specific item may be.

The target matching determination unit 140 calculates a target matching index by applying a weight to the correlation coefficient, the first similarity, the second similarity, and the third similarity according to the priority information.

The display 190 displays various similarities or target matching indices calculated by the target matching determination unit 140 on the screen.

3 is a flow chart of a target searching method based on matching possibility according to an embodiment of the present invention.

First, in the terminal including the target information database 150 storing the target face shape and body shape desired by the first user, priority information between the face shape and the body shape of the target is inputted (S310). If the priority information is set to a predefined initial value, this process (S310) may be omitted.

Next, the image of the second user is inputted (S320). In step S320, the facial image and the whole body image of the second user may be separately input, or a whole body image may be input in which the face and the body are expressed in the same manner.

When the image of the second user is inputted, the image is pre-processed to detect only the face part, and then the face shape is recognized and compared with the target face shape stored in the target information database 150 (S330).

Next, the head size is measured in the image of the second user, the ratio of each part of the body is calculated using the measured head size, the body type is classified according to the calculated ratio, And the first degree of similarity is calculated (S340).

Finally, the target matching index is calculated by applying the weights to the correlation coefficient and the first similarity according to the priority information (S370).

4 is a flowchart of a target search method based on matching possibility according to another embodiment of the present invention.

First, the terminal including the target information database 150 storing the target face shape, body shape, type and style desired by the first user receives priority information among the face shape, body shape, type and style of the target (S410 ).

Next, the image of the second user is inputted (S420).

If the second user's image is input, the image is pre-processed to detect only the face portion, and then the face shape is recognized and compared with the target face shape stored in the target information database 150 (S430).

Next, the head size is measured in the image of the second user, the ratio of each part of the body is calculated using the measured head size, the body type is classified according to the calculated ratio, And the first degree of similarity is calculated (S440).

Next, information about at least one of hair, clothes, or accessory is detected from the image of the second user, and the second similarity is calculated by comparing with the style stored in the target information database 150 (S450).

Next, the user type information of the second user is input, the questionnaire type matching test is performed, the type of the second user is classified according to the result of the matching test, the third type is compared with the target type stored in the target information database 150, The degree of similarity is calculated (S460).

Finally, the target matching index is calculated by applying the weights to the correlation coefficient, the first similarity, the second similarity, and the third similarity according to the priority information (S470).

The target matching index is a weighted sum of the correlation coefficient, the first similarity, the second similarity, and the third similarity, and can be used for measuring the liking, compatibility, and intimacy of two people.

In another embodiment of the present invention, the target matching index may be stored in a target information database or a memory of the terminal together with personal information such as the name, age, and telephone number of the second user. More preferably, when the first user searches for a name or the like in the terminal, the terminal can display the target matching index of the corresponding name.

FIG. 5 shows an example of a process of configuring a database in FIG.

First, the first user inputs a picture of a celebrity who thinks that he or she is a target, or a picture such as a photograph of himself or herself to the terminal (S501).

The terminal classifies the input image using face recognition algorithm such as Haar-like feature algorithm. The Haar-like feature algorithm extracts and simplifies the feature from the image, so that it can be classified quickly and the recognition rate is very high by using the data generated through many face learning. Eye, eyebrow, nose, mouth, jaw line, etc. from the classified face data using Haar-like feature algorithm (S502).

The portable terminal classifies the head using the image recognition algorithm in the input image, stores the ratio of the whole body using the head size, and analyzes the body shape according to the size of the body and stores it (S503). In order to more accurately set the body shape of the target, the user can separately input the whole body image of the person who is considered as the target to the portable terminal.

The fashion style information of the target obtained through the input image information or the query item is stored (S504). The fashion style can be distinguished by hair, clothes and other accessories. By applying the image recognition algorithm to the input image, it detects the accessories such as hair, clothes, and bags and analyzes the shape of hair style, accessories, and costumes by using correlation coefficient, and classifies them into suits, casuals and hip- Can be inferred. In addition, it is possible to judge whether or not an appropriate fashion style combination is determined by using hair style, costume, and accessory information, and reflect it on style information.

Next, the first user performs a matching test (entertainment disposition, personality, psychology, blood type, etc.) provided by the terminal (S505), and the terminal extracts and stores its type using data obtained through the matching test ). In this process, the first user may arbitrarily set and input the important order of each item of the face, the body, the fashion style, and the matching test.

The present invention can be implemented through software. Preferably, a program for causing a computer to execute a target search method based on matching possibility according to embodiments of the present invention may be recorded on a computer-readable recording medium and provided. When executed in software, the constituent means of the present invention are code segments that perform the necessary tasks. The program or code segments may be stored in a processor readable medium or transmitted by a computer data signal coupled with a carrier wave in a transmission medium or a communication network.

A computer-readable recording medium includes all kinds of recording apparatuses in which data that can be read by a computer system is stored. Examples of the computer readable recording medium include ROM, RAM, CD-ROM, DVD 占 ROM, DVD-RAM, magnetic tape, floppy disk, hard disk, optical data storage, and the like. In addition, the computer-readable recording medium may be distributed to network-connected computer devices so that computer-readable codes can be stored and executed in a distributed manner.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it will be understood by those skilled in the art that various changes and modifications may be made therein without departing from the spirit and scope of the invention as defined by the appended claims. It should be understood that such modifications are within the technical scope of the present invention. Accordingly, the true scope of the present invention should be determined by the technical idea of the appended claims.

Claims (13)

Receiving priority information between a face type and a body type of the target in a terminal including a target information database in which a face shape and a body shape of a target desired by the first user are stored;
Receiving an image of a second user;
Calculating a correlation coefficient by comparing a face shape of the second user with a target face shape stored in a target information database;
A head size is measured in the image of the second user, a ratio of each part of the body is calculated using the measured head size, a body type is classified according to the calculated ratio, and the body shape is compared with the target body type stored in the target information database. 1 < / RTI > And
Calculating a target matching index by applying a weight to the correlation coefficient and the first similarity according to the priority information,
Based on the matching probability. The method according to claim 1,
The face shape and the face shape of the target
Characterized in that it comprises at least one of morphological information of the eye, eyebrow, nose, mouth or chin. The method according to claim 1,
The step of calculating the first degree of similarity
Classifying the body, arms and legs in the image of the second user;
Calculating a length ratio of the body to a size of the face portion, a length ratio of the arm, and a length ratio of the legs;
Determining a body shape of the second user by comparing the body length ratio, the arm length ratio, and the leg length ratio with predefined body shape reference information, Search method. The method according to claim 1,
The step of receiving the image of the second user
And correcting the image of the second user using a gyro sensor or an acceleration sensor. 5. The method of claim 4,
The step of calculating the first degree of similarity
Measuring a distance between the terminal and the second user using an infrared sensor and calculating a ratio of each part of the body by reflecting the measured distance, Way. The method according to claim 1,
The body shape
Wherein the information is information including at least one of a long, short, seventh, seventh, seventh, seventh, seventh, seventh, seventh, seventh, seventh, seventh, seventh, seventh, seventh, seventh, seventh, seventh, seventh, The method according to claim 1,
Prior to receiving the priority information,
A ratio of each part of the body in the target image is calculated using the size of the face part detected in the target image, and the shape of the body part is calculated according to the calculated ratio, Further comprising the step of performing a matching test of the questionnaire system while receiving the user information of the first user and constructing the target information database by classifying the type of the first user according to the result of the matching test ≪ / RTI > wherein the matching is based on matching possibilities. Receiving priority information between a face shape, a body shape, and a style of the target in a terminal including a target information database in which a face shape, a body shape, and a style of a target desired by the first user are stored;
Receiving an image of a second user;
Calculating a correlation coefficient by comparing a face shape of the second user with a target face shape stored in a target information database;
A head size is measured in the image of the second user, a ratio of each part of the body is calculated using the measured head size, a body type is classified according to the calculated ratio, and the body shape is compared with the target body type stored in the target information database. 1 < / RTI >
Detecting information on at least one of hair, clothing, or accessories in the image of the second user, and comparing the style with the style stored in the target information database to calculate a second similarity; And
Calculating a target matching index by applying a weight to the correlation coefficient, the first similarity, and the second similarity according to the priority information,
Based on the matching probability. Receiving priority information of a face type, a body type, and a type of the target in a terminal including a target information database storing a face shape, a body shape and a type of a target desired by the first user;
Receiving an image of a second user;
Calculating a correlation coefficient by comparing a face shape of the second user with a target face shape stored in a target information database;
A head size is measured in the image of the second user, a ratio of each part of the body is calculated using the measured head size, a body type is classified according to the calculated ratio, and the body shape is compared with the target body type stored in the target information database. 1 < / RTI >
A third type of user is identified by comparing the type of the second user with the target type stored in the target information database, Calculating; And
Calculating a target matching index by applying a weight to the correlation coefficient, the first similarity, and the third similarity according to the priority information,
Based on the matching probability. 10. The method of claim 9,
The matching test
The method comprising: displaying the questionnaire information for any one of the first, second, third, fourth, fifth, sixth, seventh, eighth, Receiving priority information between a face type and a style of the target in a terminal including a target information database storing a face shape and a style of a target desired by the first user;
Receiving an image of a second user;
Calculating a correlation coefficient by comparing a face shape of the second user with a target face shape stored in a target information database;
Detecting information on at least one of hair, clothing, or accessories in the image of the second user, and comparing the style with the style stored in the target information database to calculate a second similarity; And
Calculating a target matching index by applying a weight to the correlation coefficient and the second similarity according to the priority information,
Based on the matching probability. 12. A computer-readable recording medium having recorded thereon a program for executing the method of any one of claims 1 to 11 in a computer system. A target information database storing a face shape and a body shape of a target desired by the first user;
Input means for receiving priority information between a face shape and a body shape of the target;
A camera for receiving an image of a second user;
A preprocessor for preprocessing an image of the second user to recognize only a facial part and then recognizing a facial shape and distinguishing an area corresponding to an arm, a leg, and a body of the second user from the image of the second user;
Calculating a correlation coefficient by comparing with a target face shape stored in the target information database, measuring a head size in the image of the second user, calculating a ratio of each part of the body using the measured head size, A similarity determination unit for comparing the body shape of the target body shape with the target body shape stored in the target information database to calculate a first degree of similarity; And
A target matching determination unit for calculating a target matching index by applying a weight to the correlation coefficient and the first similarity according to the priority information,
Wherein the matching means is based on a matching possibility.

Images