KR20080109425A - Methods and system for extracting facial features and verifying sasang constitution through image recognition - Google Patents

Abstract

A method for extracting facial features through image recognition and discriminating Sasang constitution and a system therefor are provided to diagnose the Sasang constitution of a patient remotely, objectively and accurately, thereby providing the diagnosis result to the patient who is located in a distance place. A method for extracting facial features through image recognition and discriminating Sasang constitution comprises the following steps of: receiving a face image from a user terminal(S10); removing noises from the face image and performing image smoothing(S20); converting the smoothed image of an RGB color tone space into an image of a TSL color tone space(S30); extracting a face region by using the Mahalanobis distance in the converted image(S40); extracting each feature from the extracted face region(S50); and calculating the width and proportional relation between the features(S60).

Description

Methods and System for Extracting Facial Features and Verifying Sasang Constitution through Image Recognition

1 is a block diagram illustrating a method for extracting facial features through a video recognition method and determining a constitutional structure according to the present invention;

2 is a flow chart for explaining a method for facial feature extraction and filamentous constitution determination through image recognition of the present invention;

3 is a view for explaining the distance (width) of various feature elements of the face in the method of facial feature extraction and filamentous constitution by image recognition according to the present invention.

*** Explanation of symbols for the main parts of the drawing ***

100: mobile communication terminal, 150: personal computer,

200: mobile network, 250: internet,

300: health service provider, 310: web server,

320: swap server, 330: data server,

340: an image recognition module, 350; Artificial neural network module,

360: database, 362: member information DB,

364: face image DB, 366: Sasang Constitution DB,

368: Health Information DB

The present invention relates to a method for face feature extraction and image quality determination through image recognition, and in particular, by processing a face image provided by a user, extracting feature elements constituting a face, and calculating a proportional relationship for each feature element. The present invention relates to a method of facial feature extraction and Sasang Constitution Determination through image recognition for objective and accurate discrimination of Sasang Constitution of a patient.

Recently, as interest in health care is increasing, many places provide health information. However, there are few places that determine a person's constitution and provide health information based on it. Representative methods of determining the constitution of a person is the constitution method of ideology. As is well known, ideology is a field of oriental medicine that divides the human constitution into four, created by Ezema. He was written in 1894 by the Sujebobobo, which he wrote, and human beings have a wealth of books, and because of this, the affection of noble love works. He caused physiological phenomena and insisted on proper food and curing methods.

This ideology of ideology is the core of constitution medicine, and knowing each constitution can be a shortcut to preventing and treating diseases. The four constitutions of ideology are the Sunin, Taiyangin, Taeumin, and Soinin. It is said that when something more false is more false or something more poor, it appears as a disease, and this is called Tae-Soo. Here, Tae-so is not an anatomical meaning, but refers to a function, the sun person is a lung liver liver (肺 大 肝 小), Soyangin a non-large kidney (脾 大 腎 小), Taeumin is a liver lung (肝 大 肺 小), Suu Yin was called the New Daeso (腎 大 脾 小).

On the other hand, there are methods for determining filamentous constitutions, such as the appearance, the mentality, and the symptom. Symptoms were determined by questionnaire.

However, in the conventional method for determining Sasang constitution, the appearance of the method is not only inconvenient for the patient to visit the hospital, but also leads to a different result from the constitution. There was a problem that could be. In addition, even in the case of a questionnaire, the subjective view of the patient may be reflected in an ambiguous state, and according to the answer, there is a problem in that a discriminant result different from the actual constitution may be derived.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned problems, and after processing feature images provided by a user to extract feature elements constituting the face, calculating proportional relations for each feature element to remotely control the ideological constitution of the patient. Also aims to provide a method for facial feature extraction and filamentous constitution by using image recognition to objectively and accurately discriminate and to provide health information suitable for such discrimination results.

The facial feature extraction and Sasang constitution determination method through the image recognition of the present invention for achieving the above object is performed by the health service provider in a state in which the health service provider is connected to the user terminal through a wired / wireless communication network, (A) receiving a face image from a user terminal; (B) performing noise reduction and image smoothing on the face image received in step (a); (C) converting an image of the RGB smoothed color space into an image of the TSL color space in the step (b); Extracting a facial region using the Mahalanobis distance from the image converted in step (c), and extracting each feature element from the facial region extracted in step (d), and then (E) calculating the width and its proportionality.

In the above configuration, it is preferable to further include the step (f) of determining the filamentous constitution for the user by processing the proportional relationship calculated in the step (e) with an artificial neural network algorithm.

The feature element may be any one or more of nasal width, inner eye width, intraocular width, rectangle width, outer eye width, non-wing width or eye width, and the proportional relationship may be the width-to-person width.

The method may further include the step (g) of providing the user with health information suitable for the filamentous constitution determined in the step (f).

In addition, the apparatus for performing the facial feature extraction and image constitution determination method through the image recognition, the health connected to a mobile communication terminal or a personal computer with a function capable of transmitting a picture through a mobile communication network or the Internet network wired and wireless In the service providing server,

The health service providing server includes a web server for processing various web documents with the personal computer, a swap server for processing various swap documents with the mobile communication terminal, and data for inquiring various data. An image recognition module for extracting facial feature elements by performing image recognition on a face photograph of a user transmitted from a server, a remote mobile communication terminal or a personal computer, and calculating various proportional relations between the extracted facial features, and the image recognition It comprises an artificial neural network module for processing various proportional relation data between facial feature elements calculated by the module with an artificial neural network algorithm to determine the most suitable filamentous constitution, and a database containing various necessary data.

In the present invention, the database is a member information DB, including a member ID and password, gender, age, address and phone number, etc. for the member information, for example, the original image for the face image provided from the user According to the Sasang Constitution DB and Sasang Constitution, which stores the face image DB, various data related to the user's Sasang Constitution, that is, the width and proportional relationship of the facial feature elements of the user and the Sasang Constitution information determined by the user It includes various health information, that is, a health information DB containing foods, medicines, or prescriptions suitable for each constitution.

Hereinafter, with reference to the accompanying drawings, it will be described in detail a preferred embodiment of the facial feature extraction and filamentous constitution determination method and apparatus through image recognition of the present invention.

1 is a diagram illustrating a network configuration of a method for extracting facial features and determining a constitutional structure through image recognition according to the present invention. As shown in FIG. 1, the network configuration in which the facial feature extraction and the image constitution determination method through the image recognition of the present invention are implemented is largely a mobile communication terminal 100 or a personal computer 150 (hereinafter, collectively referred to as' A user terminal 'may be connected to a health service provider or a health service providing server 300 through a wired / wireless communication network such as a mobile communication network 200 or the Internet 250.

In the above-described configuration, the mobile communication terminal 100 may be a mobile phone equipped with a function of transmitting a picture to the wireless Internet, that is, the health service providing server 300, and furthermore, a digital camera is attached to the picture taking function. It is desirable to have a mobile phone with this. Of course, the mobile communication terminal 100 may be a WiBro terminal having a portable Internet function. The personal computer 150 may be any one as long as it can transmit a picture to the Internet function, that is, the health service providing server 300.

Meanwhile, the health service providing server 300 may perform various WAP documents between the web server 310 and the mobile communication terminal 100, which are in charge of processing various web documents, with the personal computer 150. The user's face received from the swap server 320 in charge of the processing, the data server 330 for managing a database to be described later and querying various data therefrom, the mobile communication terminal 100 or the personal computer 340 at a remote location. Image recognition module 340 for extracting facial feature elements by performing image recognition on a photo and calculating various proportional relations between the extracted facial features, and various facial feature elements calculated by the image recognition module 340. The artificial neural network module 350 for processing the proportional relationship data with an artificial neural network algorithm to determine the most suitable filamentous constitution and a database 360 for storing various necessary data It can be achieved by also.

The database 360 includes member information DB 362 containing various information about the member, for example, a member ID and password, gender, age, address, and telephone number, and an original image of a face image provided from the user. The facial image DB 364 storing the user's cadastral constitution, and the cadastral constitution DB storing various data related to the user's thought constitution, that is, the width and proportional relationship of the facial feature elements of the user and the thought constitution information determined by the user. 366) and a variety of health information according to Sasang constitutions, that is, health information DB 368 containing foods, medicines, or prescriptions suitable for each constitution.

FIG. 2 is a flowchart illustrating a method of extracting facial features through a video recognition method and determining a ideological constitution, which is performed by a health service provider or a health service providing server 300. As shown in FIG. 2, in step S10, the health service provider 300 receives a face image from a user through a web server 310 or a swap server 320 and stores the face image in the face image DB 364. Next, the image recognition module 340 of the health service provider 300 performs preprocessing, specifically noise removal and image smoothing, on the face image of the user thus received in step S20.

Next, in step S30, color conversion is performed in order to extract the face area more accurately, that is, the color space of RGB (Red, Green, Yellow) received from the user is converted into TSL (Tint, Saturation, Lightness) color space. In this TSL color space, the TL plane has a strong advantage in color distribution and the influence of lighting. As described above, in the present invention, the TSL color space is used, and the equation for changing the RGB color space into the TSL color space is shown in Equation 1 below.

      In Equation 1 above

,

,

,

이다.

,

,

,

to be.

Next, in step S40, the face area is extracted by the Mahalanobis distance as shown in Equation 2 below using the color of the center pixel under the assumption that the face area is in the center of the picture.

In Equation 2 above, X is an input image, K is a face region in the side image, C _k is a covariance of K and m _k is a response value for T and S colors. On the other hand, Mahalanobis distance is the most widely used distance concept in cluster analysis. It is characterized by not only the simple distance between two points but also the standard deviation and correlation coefficient that represent the characteristics of variables. In cluster analysis, most variables are normalized to transformed variables with a mean of 0 and a variance of 1 before clustering. The Mahalanobis distance is a distance created by considering the standard deviation and correlation between variables. If all variables are standardized and all variables are independent of each other, the Mahalanobis distance is equal to the Euclidean distance. Will match.

When the face region is extracted as described above, in step S50, the small region is removed using the open operation and the close operation of the shape processing, and the background region is removed again using labeling, and then the positional relationship of each feature is used. Various feature elements such as eyebrows, eyes, nose, lips and facial contours are extracted. Here, Morphological processing means changing the form while maintaining the basic characteristics of the image. Basically, there are erosion and dilation, and opening using both. And Closing operations.

Next, in step S60, the distance (width) between the extracted feature elements and their proportional relationship are calculated.

3 is a view for explaining the distance (width) of various feature elements of the face in the method of facial feature extraction and filamentous constitution by image recognition according to the present invention. As shown in FIG. 3, in the facial feature extraction and filamentous constitution method using image recognition according to the present invention, the nasal width (c), the inner eye width (b), the maximum eye width (g), the interval width (f), etc. The proportional relationship obtained by dividing the angle width (a), the non-wing width (d), the eye width, and the like by the width (e) of the weight is used. In other words, among the facial features, the nasal width ratio, inner eye width ratio, maximum eye width ratio, rectangle width ratio, outer eye angle width ratio, non-wing width ratio and eye width ratio are used among face features. First, filamentous constitution discrimination criteria based on nasal width, ie, nasal width ratio (x ₁ = c / e * 100) representing the ratio of weight to root of nose, are shown in Table 1 below.

section constitution x ₁ <= 139.7 Soyangin 139.7 <= x ₁ <155.32 Noisemaker, Soyangin 155.32 <= x ₁ <156.12 Soyangin 156.12 <= x ₁ <156.76 Soyangin, Taein x ₁ > 156.76 Lunar

Next, filamentous constitution discrimination criteria based on the inner eye width ratio (i.e., the inner eye width ratio (x ₂ = b / e * 100) representing the ratio of the weight of the person to the width between the inside of the eye) are shown in Table 2 below. same.

section constitution x ₂ <= 291.83 Noisemaker 291.83 <= x ₂ <315.53 Noisemaker, Soyangin 315.53 <= x ₂ <322.11 Soyangin 322.11 <= x ₂ <326.30 Soyangin, Taein x ₂ > 326.30 Lunar

Next, filamentous constitution discrimination criteria based on the maximum width (顔 大 大 幅), that is, the maximum width ratio (x ₃ = g / e * 100) representing the ratio of the weight to the face width is shown in Table 3 below.

section constitution x ₃ <= 1181.31 Noisemaker 1181.31 <= x ₃ <1247.40 Noisemaker, Soyangin 1247.40 <= x ₃ <1276.94 Soyangin 1276.94 <= x ₃ <1314.25 Soyangin, Taein x ₃ > 1314.25 Lunar

Next, the criteria for discriminating filamentous constitution by the interval width ratio (x ₄ = f / e * 100) representing the width width of the corner, that is, the authentication width ratio to the width of the mouth, are shown in Table 4 below.

section constitution x ₄ <= 403.17 Noisemaker 403.17 <= x ₄ <415.27 Noisemaker, Soyangin 415.27 <= x ₄ <435.09 Soyangin 435.09 <= x ₄ <444.70 Soyangin, Taein x ₄ > 444.70 Lunar

Next, filamentous constitution discrimination criteria based on the external angle width ratio, i.e., the external angle width ratio (x ₅ = a / e * 100) representing the authentication width ratio to the outside width of the two eyes are shown in Table 5 below. .

section constitution x ₅ <= 731.94 Noisemaker 31.94 <= x ₅ <776.13 Noisemaker, Soyangin 776.13 <= x ₅ <789.04 Soyangin 789.04 <= x ₅ <814.97 Soyangin, Taein x> 814.97 Lunar

Next, filamentous constitution determination criteria based on the non-wing width, that is, the non-wing width ratio (x ₆ = d / e * 100) representing the authentication width ratio to the nose wing width, are shown in Table 6 below.

section constitution x ₆ <= 318.87 Noisemaker 318.87 <= x ₆ <331.61 Noisemaker, Soyangin 331.61 <= x ₆ <347.52 Soyangin 347.52 <= x ₆ <356.21 Soyangin, Taein x ₆ > 356.21 Lunar

Finally, the criteria for judging constitution by the depth ratio (x ₇ = a-b / e * 100) representing the authentication width ratio to the depth of eye width are shown in Table 7 below.

section constitution x ₇ <= 438.18 Noisemaker 438.18 <= x ₇ <462.28 Noisemaker, Soyangin 462.28 <= x ₇ <464.31 Soyangin 464.31 <= x ₇ <490.62 Soyangin, Taein x ₇ > 490.62 Lunar

2, in step S70, the result of calculating the proportions between the facial feature elements is input to the artificial neural network module 350 to finally determine the filamentous constitution. The artificial neural network algorithm simplifies the process of neurotransmission in living organisms and mathematically interprets the model. The neuron connection strength is passed through the intertwined neurons (minimal units of a network in neural networks). Analyze problems through a kind of training process that regulates This process is similar to the process of learning and remembering human beings, which enables reasoning, classification, and prediction. Currently, artificial neural networks are used for optimization problems and prediction problems.

Finally, in step S80, various health information such as health information prescription suitable for the user is provided to the user based on the determined constitution.

The facial feature extraction and filament constitution determination method through image recognition of the present invention are not limited to the above-described embodiments, and various modifications can be carried out within the range allowed by the technical idea of the present invention.

According to the method of facial feature extraction and filamentous constitution according to the present invention as described above, after extracting the feature elements constituting the face by processing the face image provided from the user, the proportional relationship of each feature element is determined. By calculating, the filamentous constitution of the patient can be determined objectively and accurately even at a remote location, and furthermore, health information suitable for this determination result can be provided at a remote location.

Claims (6)

Performed by the health care provider while the health care provider is connected to the user terminal through a wired or wireless network, (A) receiving a face image from a user terminal; (B) performing noise reduction and image smoothing on the face image received in step (a); (C) converting an image of the RGB smoothed color space into an image of the TSL color space in the step (b); (D) extracting a face region using the Mahalanobis distance from the image converted in step (c); and (E) calculating the width and the proportional relationship between the feature elements after extracting each feature element from the face region extracted in step (d). Way. The method of claim 1, And (f) determining the filamentous constitution for the user by processing the proportional relationship calculated in the step (e) using an artificial neural network algorithm. . The method of claim 2, The feature element is any one or more of the nasal width, the inner eye width, the maximum eye width, the inter-angle width, the outer eye width, nasal width or eye width, And the proportional relationship is the width versus the width of the face. The method according to any one of claims 1 to 3, And (g) providing the user with health information suitable for the filamentous constitution determined in the step (f). In the health service providing server connected to a mobile communication terminal or a personal computer equipped with a function for transmitting a picture through a mobile communication network or the Internet via wired or wireless, The health service providing server includes a web server for processing various web documents with the personal computer, a swap server for processing various swap documents with the mobile communication terminal, and data for inquiring various data. An image recognition module for extracting facial feature elements by performing image recognition on a face photograph of a user transmitted from a server, a remote mobile communication terminal or a personal computer, and calculating various proportional relations between the extracted facial features, and the image recognition An artificial neural network module for processing various proportional relationship data between facial feature elements calculated by the module by an artificial neural network algorithm to determine the most suitable filamentous constitution, and an image including a database containing various necessary data. Facial Feature Extraction and Sasang Constitution through Recognition Discrimination device. The method of claim 5, The database is a face image that stores various information about a member, for example, a member information DB including a member ID and password, gender, age, address, and phone number, and an original image of a face image provided by a user. DB, various data related to the user's Sasang constitution, that is, the Sasang Constitution DB and various health information according to the Sasang Constitution which store the width and proportional relationship of the facial feature elements of the user and the Sasang Constitution information determined by the user Apparatus for extracting facial features and filamentous constitution by image recognition, comprising a health information DB containing foods, medicines, or prescriptions suitable for each constitution.

Images