Traditional Chinese medicine inspection diagnosis face division method and system based on big data

Abstract

The invention discloses a traditional Chinese medicine inspection diagnosis face division method and system based on big data. According to the method, after the face is preprocessed based on the theory of traditional Chinese medicine, a modular face inspection division model is constructed, the face is divided into different faces through the trained model according to the preprocessed and divided modular face image features, and the face image is divided based on the theory of traditional Chinese medicine. The face is subjected to color observation part division or qi observation part division. According to the method, the misdiagnosis rate caused by a subjective diagnosis method of color observation is reduced through the established modular face inspection division model, and meanwhile, the efficiency of targeted color observation and air observation face division diagnosis is improved by utilizing the improved model on the basis of traditional Chinese medicine big data.

Description

Technical Field

[0001] This invention belongs to the field of intelligent medical image processing technology, and in particular relates to a method and system for facial segmentation in traditional Chinese medicine based on big data observation. Background Technology

[0002] The four diagnostic methods of Traditional Chinese Medicine (TCM) are observation, auscultation, inquiry, and palpation. Observation refers to observing the complexion; auscultation to listening to sounds; inquiry to asking about symptoms; and palpation to feeling the pulse. These are collectively known as the four diagnostic methods. Observation is the first of the four, and its main components are face diagnosis and tongue diagnosis.

[0003] Facial observation refers to assessing the health of a person and their internal organs by observing changes in the complexion and color of their facial features, thus providing a basis for diagnosis and treatment. Traditional Chinese medicine views the human body as an organic whole, considering the five internal organs as the center, meridians as transmission channels, and qi and blood as the medium connecting the organs, tissues, skin, senses, and limbs. The face and internal organs are seen as having an invisible reflection; by observing changes in the complexion and color of the facial features, one can indirectly understand the condition of the five internal organs.

[0004] Dong Mengqing et al. used a digital detection instrument for TCM facial diagnosis to analyze the facial color characteristics of 259 patients with chronic renal failure, 128 patients with coronary heart disease, and 232 patients with chronic hepatitis B. The results showed that there were significant differences in facial color index among the three diseases.

[0005] The existing TCM facial diagnosis equipment - the TCM Four Diagnosis Instrument - integrates a large number of modern technological achievements and the clinical experience of many TCM experts, combining TCM tongue diagnosis, facial diagnosis, pulse diagnosis and questioning. The TCM Four Diagnosis Instrument facial diagnosis requires the acquisition of a facial image, which includes the face, hair and environmental background. It is necessary to extract the face part of the image or remove the hair and environmental background.

[0006] The drawbacks of existing technology are:

[0007] 1. Nowadays, traditional Chinese medicine diagnoses patients' conditions by subjectively observing their facial color. However, since the doctor's observation of facial color is subjective, inaccurate diagnostic results may occasionally occur.

[0008] 2. In addition, existing facial diagnostic models mostly use multiple models to process image features to obtain an overall facial segmentation image, which makes the detection process cumbersome and the equipment structure complex.

[0009] 3. In existing technologies, facial segmentation images only divide the face into areas such as the Mingtang, Que, Ting, Fan, and Bi, rather than using the five-color diagnosis method in traditional Chinese medicine to differentiate between observing the qi or observing the color of the face, resulting in inaccurate diagnosis.

[0010] Therefore, it is necessary to redesign a traditional Chinese medicine facial diagnosis method and system based on big data to address the existing problems. Summary of the Invention

[0011] To address the aforementioned technical problems, this invention proposes a method and system for facial segmentation in traditional Chinese medicine based on big data observation.

[0012] In a first aspect of the present invention, a method for facial segmentation in traditional Chinese medicine based on big data observation is provided, characterized by comprising the following steps:

[0013] A1. Obtain the first color image of the face;

[0014] A2. Based on the preprocessing of the first face color image, a first dark module image and a first light module image are obtained, and the first dark region face segmentation method and the first light region face segmentation method are respectively acquired from the first dark module image and the first light module image.

[0015] A3. Construct a modular facial diagnosis segmentation model based on the first input features composed of the first dark module image and the first light module image, and the first output features composed of the first dark region facial segmentation method and the first light region facial segmentation method.

[0016] A4. Receive the second face color image and generate a second output feature composed of the second dark region face segmentation method and the second light region face segmentation method according to the modular facial diagnosis segmentation model.

[0017] A5. Based on the second output feature composed of the second dark area facial segmentation method and the second light area facial segmentation method, the face is divided into areas for visual observation or areas for visual assessment, assisting doctors in making visual diagnosis.

[0018] Furthermore, before the preprocessing of the first face color image to obtain the first dark module image and the first light module image, the first face color image or the second face color image is divided into five colors—green, red, yellow, white, and black—according to the five-color diagnosis theory of traditional Chinese medicine. Red and black are designated as dark colors, and green, white, and yellow are designated as light colors. The color gamut feature value P of the dark color image is set as follows. d The calculation formula and the color gamut feature value P set as light color. L .

[0019] Furthermore, the modular facial diagnosis segmentation model is constructed using deep learning methods.

[0020] Furthermore, red and black are defined as dark colors, cyan, white and yellow are defined as light colors, and the variance F of the color intensity values ​​between dark and light colors is defined.

[0021] Furthermore, the color intensity value variance F is used to distinguish between dark and light colors. Regions where the average intensity value of the RGB three color channels in the first or second face color image is greater than the color intensity value variance F are considered light-colored regions, and regions where the average intensity value of the RGB three color channels in the first or second face color image is less than the color intensity value variance F are considered dark-colored regions. The color gamut feature value P of the light-colored region... L And the color gamut characteristic value P of dark colors d The input features P are obtained by combining them, and the vector representation of P is P(P L P d ).

[0022] Furthermore, the deep learning method employs an improved convolutional neural network.

[0023] A facial segmentation system based on big data in traditional Chinese medicine is also provided, characterized in that the system includes:

[0024] Color face image acquisition module: acquires a first color face image; also acquires a second color face image;

[0025] Facial image color preprocessing module: Based on the preprocessing of the first face color image, a first dark module image and a first light module image are obtained, and the first dark region facial segmentation method and the first light region facial segmentation method are respectively acquired from the first dark module image and the first light module image.

[0026] Modular facial diagnosis segmentation model construction module: Construct a modular facial diagnosis segmentation model based on the first input features composed of the first dark module image and the first light module image, and the first output features composed of the corresponding first dark region facial segmentation method and the first light region facial segmentation method;

[0027] The facial segmentation processing module for visual diagnosis receives a second color image of a face and generates a second output feature composed of a second dark region facial segmentation method and a second light region facial segmentation method according to the modular facial visual diagnosis segmentation model. Based on the second output feature composed of the second dark region facial segmentation method and the second light region facial segmentation method, the module divides the face into areas for visual diagnosis or areas for visual aura analysis to assist doctors in making visual diagnosis judgments.

[0028] Furthermore, before the preprocessing of the first face color image to obtain the first dark module image and the first light module image, the first face color image or the second face color image is divided into five colors—green, red, yellow, white, and black—according to the five-color diagnosis theory of traditional Chinese medicine. Red and black are designated as dark colors, and green, white, and yellow are designated as light colors. The color gamut feature value P of the dark color image is set as follows. d And the color gamut feature value P set to light color L .

[0029] Furthermore, red and black are defined as dark colors, cyan, white and yellow are defined as light colors, and the variance F of the color intensity values ​​between dark and light colors is defined.

[0030] The color intensity variance F is used to distinguish between dark and light colors. Regions where the average intensity value of the RGB channels in either the first or second face color image is greater than the color intensity variance F are considered light regions, and regions where the average intensity value of the RGB channels is less than the color intensity variance F are considered dark regions. The color gamut feature value P of the light regions is... L And the color gamut characteristic value P of dark colors d The input features P are obtained by combining them, and the vector representation of P is P(P L P d ).

[0031] Furthermore, the modular facial diagnosis segmentation model is constructed using a deep learning method, which employs an improved convolutional neural network.

[0032] This invention not only utilizes the subjective observation of a patient's facial color in Traditional Chinese Medicine (TCM) but also incorporates image processing technology to segment the patient's face for diagnosis, reducing the likelihood of diagnostic errors. Furthermore, the facial diagnosis model employs image feature preprocessing based on the TCM five-color theory to obtain an overall facial segmentation image before comprehensive facial segmentation, reducing the complexity of subsequent feature recognition. Only a client capable of image recognition needs to process the data to obtain a facial segmentation method for different areas of the face, thus more accurately distinguishing between observing the "qi" (vital energy) and observing the "color" (facial complexion) in TCM's five-color diagnosis. The facial feature segmentation required for TCM facial diagnosis utilizes a convolutional neural network with an improved activation function, making the model based on large-scale facial data processing more accurate and reliable.

[0033] Further embodiments and improvements of the present invention will be described in conjunction with the accompanying drawings and specific examples. Attached Figure Description

[0034] Figure 1 This is a flowchart of a traditional Chinese medicine facial segmentation method based on big data observation according to the present invention;

[0035] Figure 2 This is a schematic diagram of a traditional Chinese medicine facial segmentation system based on big data observation according to the present invention;

[0036] Figure 3 This is a schematic diagram illustrating the division of dark and light regions in an embodiment of the present invention;

[0037] Figure 4 This is a schematic diagram illustrating the basic principle of the convolutional neural network in the embodiments of this invention;

[0038] Figure 5 This is a schematic diagram of an electronic device structure for implementing the method of the present invention in an embodiment of the present invention. Detailed Implementation

[0039] The invention will now be further described in conjunction with the accompanying drawings and specific embodiments.

[0040] To address the aforementioned technical problems, this invention provides a method and system for facial segmentation in traditional Chinese medicine based on big data observation.

[0041] In a first aspect of the present invention, a method for facial segmentation in traditional Chinese medicine based on big data observation is provided, characterized by comprising the following steps:

[0042] A1. Obtain the first color image of the face;

[0043] A2. Based on the preprocessing of the first face color image, a first dark module image and a first light module image are obtained, and the first dark region face segmentation method and the first light region face segmentation method are respectively acquired from the first dark module image and the first light module image.

[0044] A3. Construct a modular facial diagnosis segmentation model based on the first input features composed of the first dark module image and the first light module image, and the first output features composed of the first dark region facial segmentation method and the first light region facial segmentation method.

[0045] A4. Receive the second face color image, and generate a second output feature composed of the second dark region face segmentation method and the second light region face segmentation method according to the modular facial diagnosis segmentation model.

[0046] A5. Based on the second output feature composed of the second dark area facial segmentation method and the second light area facial segmentation method, the face is divided into areas for visual observation or areas for visual assessment, assisting doctors in making visual diagnosis.

[0047] Furthermore, before the preprocessing of the first face color image to obtain the first dark module image and the first light module image, the first face color image or the second face color image is divided into five colors—green, red, yellow, white, and black—according to the five-color diagnosis theory of traditional Chinese medicine. Red and black are designated as dark colors, and green, white, and yellow are designated as light colors. The color gamut feature value P of the dark color image is set as follows. d The calculation formula and the color gamut feature value P set as light color. L The calculation formula is:

[0048]

[0049]

[0050] In the formula, k1, k2, and k3 are the gamut weight values ​​for the three color channels RGB, R, G, and B are the intensity characteristic values ​​of the red, green, and blue color channels, and S... d S represents the area of ​​the face occupied by dark colors. L Let S represent the area of ​​the face occupied by the light-colored area. In this embodiment, the values ​​of k1, k2, and k3 are 0.34, 0.36, and 0.40, respectively.

[0051] The face area is obtained through a general face recognition system, where the dark and light areas are obtained after preprocessing and segmentation.

[0052] In this embodiment, the values ​​of R, G, and B range from 0 to 255. The RGB values ​​of pure cyan are (0, 255, 255), while the values ​​of G and B generally fluctuate within a certain range. The RGB values ​​of pure red are (255, 0, 0). If we distinguish them by pure color, we will miss the corresponding color system image features. Therefore, in this embodiment, we only distinguish between dark and light colors based on the variance F of the color intensity value.

[0053] Furthermore, the modular facial diagnosis segmentation model is constructed using deep learning methods.

[0054] Furthermore, red and black are defined as dark colors, and cyan, white, and yellow are defined as light colors. The variance F of the color intensity values ​​between dark and light colors is defined as follows:

[0055]

[0056] In the formula, g is the average of the RGB three color channel intensity values ​​of the first face color image or the second face color image, and is the average of the RGB three color channel intensity values ​​of the first face color image or the second face color image.

[0057] g0

[0058] The average value of the R intensity value of the color channel in the color image is either the first color image of the face or the second color image.

[0059] g1

[0060] Mean intensity values ​​of the G color channel in a color image of a human face.

[0061] Furthermore, the color intensity value variance F is used to distinguish between dark and light colors. Regions where the average intensity value of the RGB three color channels in the first or second face color image is greater than the color intensity value variance F are considered light-colored regions, and regions where the average intensity value of the RGB three color channels in the first or second face color image is less than the color intensity value variance F are considered dark-colored regions. The color gamut feature value P of the light-colored region... L And the color gamut characteristic value P of dark colors d The input features P are obtained by combining them, and the vector representation of P is P(P L P d ).

[0062] Activation functions are an extremely important feature of artificial neural networks. They determine whether a neuron should be activated. Activation means that the information received by the neuron is related to the given information. Activation functions perform nonlinear transformations on the input information and then pass the transformed output information as input information to the next layer of neurons. Therefore, activation functions are very important to neural networks.

[0063] Furthermore, the deep learning method employs an improved convolutional neural network, and the calculation formula for its activation function m(x) is shown below:

[0064]

[0065] x is the first input feature or the second input feature, S d S represents the area of ​​the face occupied by dark colors. L P represents the area of ​​the face occupied by lighter colors. d P represents the gamut characteristic value of dark colors. L The color gamut feature value is for light colors.

[0066] The output features derived from the convolutional neural network model with the improved activation function in this embodiment are used to classify facial features into areas for visual observation or areas for visual assessment, assisting doctors in making diagnostic judgments through observation.

[0067] In this embodiment, the corresponding dark region face segmentation method and light region face segmentation method are performed based on the output features of the final improved convolutional neural network output layer. Specifically, the final output layer vector is (0, 1) or (1, 0) or (0, 0) or (1, 1). The output layer vector is (0, 1) or (1, 0), which means that the dark region face segmentation method is used to segment the face by color. If the final output layer vector is (0, 0), the face needs to be segmented by aura. If the final output layer vector is (1, 1), the face needs to be segmented by both aura and color.

[0068] A facial segmentation system based on big data in traditional Chinese medicine is also provided, characterized in that the system includes:

[0069] Color face image acquisition module: acquires a first color face image; also acquires a second color face image;

[0070] Facial image color preprocessing module: Based on the preprocessing of the first face color image, a first dark module image and a first light module image are obtained, and the first dark region facial segmentation method and the first light region facial segmentation method are respectively acquired from the first dark module image and the first light module image.

[0071] Modular facial diagnosis segmentation model construction module: Construct a modular facial diagnosis segmentation model based on the first input features composed of the first dark module image and the first light module image, and the first output features composed of the corresponding first dark region facial segmentation method and the first light region facial segmentation method;

[0072] The facial segmentation processing module for visual diagnosis receives a second color image of a face and generates a second output feature composed of a second dark region facial segmentation method and a second light region facial segmentation method according to the modular facial visual diagnosis segmentation model. Based on the second output feature composed of the second dark region facial segmentation method and the second light region facial segmentation method, the module divides the face into areas for visual diagnosis or areas for visual aura analysis to assist doctors in making visual diagnosis judgments.

[0073] Furthermore, before the preprocessing of the first face color image to obtain the first dark module image and the first light module image, the first face color image or the second face color image is divided into five colors—green, red, yellow, white, and black—according to the five-color diagnosis theory of traditional Chinese medicine. Red and black are designated as dark colors, and green, white, and yellow are designated as light colors. The color gamut feature value P of the dark color image is set as follows. d The calculation formula and the color gamut feature value P set as light color. L The calculation formula is:

[0074]

[0075]

[0076] In the formula, k1, k2, and k3 are the gamut weight values ​​for the three color channels RGB, R, G, and B are the intensity characteristic values ​​of the red, green, and blue color channels, and S... d S represents the area of ​​the face occupied by dark colors. L The area represented by the light color is the face area, and S represents the face area.

[0077] Furthermore, red and black are defined as dark colors, and cyan, white, and yellow are defined as light colors. The variance F of the color intensity values ​​between dark and light colors is defined as follows:

[0078]

[0079] In the formula, g is the average of the RGB three color channel intensity values ​​of the first face color image or the second face color image, and is the average of the RGB three color channel intensity values ​​of the first face color image or the second face color image.

[0080] g0

[0081] The average value of the R intensity value of the color channel in the color image is either the first color image of the face or the second color image.

[0082] g1

[0083] Mean G intensity values ​​of the color channels in a color image of a human face;

[0084] The color intensity variance F is used to distinguish between dark and light colors. Regions where the average intensity value of the RGB channels in either the first or second face color image is greater than the color intensity variance F are considered light regions, and regions where the average intensity value of the RGB channels is less than the color intensity variance F are considered dark regions. The color gamut feature value P of the light regions is... L And the color gamut characteristic value P of dark colors d The input features P are obtained by combining them, and the vector representation of P is P(P L P d ).

[0085] Furthermore, the modular facial diagnosis segmentation model is constructed using a deep learning method. This deep learning method employs an improved convolutional neural network, and the calculation formula for its activation function m(x) is shown below:

[0086]

[0087] x is the first input feature or the second input feature, S d S represents the area of ​​the face occupied by dark colors.L P represents the area of ​​the face occupied by lighter colors. d P represents the gamut characteristic value of dark colors. L The color gamut feature value is for light colors.

[0088] This invention not only utilizes the subjective observation of a patient's facial color in Traditional Chinese Medicine (TCM) but also incorporates image processing technology to segment the patient's face for diagnosis, reducing the likelihood of diagnostic errors. Furthermore, the facial diagnosis model employs image feature preprocessing based on the TCM five-color theory to obtain an overall facial segmentation image before comprehensive facial segmentation, reducing the complexity of subsequent feature recognition. Only a client capable of image recognition needs to process the data to obtain a facial segmentation method for different areas of the face, thus more accurately distinguishing between observing the "qi" (vital energy) and observing the "color" (facial complexion) in TCM's five-color diagnosis. The facial feature segmentation required for TCM facial diagnosis utilizes a convolutional neural network with an improved activation function, making the model based on large-scale facial data processing more accurate and reliable.

[0089] For any module structures not specifically defined in this invention, the existing technical specifications shall prevail. The existing technical specifications mentioned in the foregoing background and specific embodiments sections are considered part of this invention and are used to understand the meaning of certain technical features or parameters. The scope of protection of this invention is determined by the actual contents of the claims.

Claims

1. A facial segmentation method for TCM inspection based on big data, characterized in that: The following steps are involved: A1, obtaining a first face color image; A2, obtaining a first dark module image and a first light module image according to the preprocessing division performed on the first face color image, and collecting a first dark area facial division method and a first light area facial division method of the first dark module image and the first light module image respectively; A3, constructing a modular facial inspection segmentation model according to a first input feature composed of the first dark module image and the first light module image and a first output feature composed of the corresponding first dark area facial segmentation method and the first light area facial segmentation method; A4, receiving a second human face color image, and generating second output features consisting of a second dark area facial segmentation method and a second light area facial segmentation method according to the modular facial inspection segmentation model; A5. Divide the face into color-observing areas or qi-observing areas according to the second output feature composed of the second dark-color area facial segmentation method and the second light-color area facial segmentation method, so as to assist the doctor in making visual diagnosis.

2. A method for facial segmentation based on TCM inspection in big data as claimed in claim 1, characterized in that: Before the preprocessing division performed on the first face color image to obtain the first dark module image and the first light module image, the first face color image or the second face color image is divided into five colors, namely, cyan, red, yellow, white and black, according to the five-color diagnosis theory of traditional Chinese medicine, and the red and black are set as dark colors, cyan, white and yellow are set as light colors, and the color gamut feature value P of the dark color is set. d And the color gamut feature value P set to light color L .

3. The method for facial segmentation based on TCM inspection in big data as claimed in claim 1, characterized in that: The modular facial inspection and segmentation model is constructed using a deep learning method.

4. A method for facial segmentation based on TCM inspection in big data as claimed in claim 2, characterized in that: The red and black are set as dark colors, the cyan, white and yellow are set as light colors, and the color intensity value variance F between the dark colors and the light colors is defined.

5. A method for facial segmentation based on TCM inspection in big data as claimed in claim 4, characterized in that: The color intensity value variance F is used to distinguish between dark and light colors, and the portion where the mean of the intensity values ​​of the three color channels RGB in the first face color image or the second face color image is greater than the color intensity value variance F is a light area, and the portion where the mean of the intensity values ​​of the three color channels RGB in the first face color image or the second face color image is less than the color intensity value variance F is a dark area. The color gamut feature value P of the light area L And the dark color gamut feature value P d The input feature P is combined and the vector of P is represented as P(P L , P d ).

6. The method for facial segmentation based on TCM inspection in big data as claimed in claim 3, characterized in that: The deep learning method adopts an improved convolutional neural network.

7. A TCM facial segmentation system based on big data, characterized in that: The system comprises: Color face image acquisition module: acquires a first face color image; also acquires a second face color image; A facial image dark and light color preprocessing module is used to obtain a first dark module image and a first light module image according to the preprocessing division of the first facial color image, and collect a first dark area facial division method and a first light area facial division method of the first dark module image and the first light module image respectively; A modular facial inspection segmentation model construction module: constructing a modular facial inspection segmentation model according to a first input feature consisting of the first dark module image and the first light module image and a first output feature consisting of the corresponding first dark area facial segmentation method and the first light area facial segmentation method; The visual diagnosis facial segmentation processing module receives a second facial color image, and generates a second output feature consisting of a second dark area facial segmentation method and a second light area facial segmentation method according to the modular facial visual diagnosis segmentation model; divides the face into color-observation areas or qi-observation areas according to the second output feature consisting of the second dark area facial segmentation method and the second light area facial segmentation method, to assist the doctor in making visual diagnosis judgments.

8. A TCM facial segmentation system based on big data as claimed in claim 7, characterized in that: Before the preprocessing division performed on the first face color image to obtain the first dark module image and the first light module image, the first face color image or the second face color image is divided into five colors, namely, cyan, red, yellow, white and black, according to the five-color diagnosis theory of traditional Chinese medicine, and the red and black are set as dark colors, cyan, white and yellow are set as light colors, and the color gamut feature value P of the dark color is set. d And the color gamut feature value P set to light color L .

9. A TCM facial segmentation system based on big data as claimed in claim 8, characterized in that: The red and black are set as dark colors, the cyan, white and yellow are set as light colors, and the color position intensity value variance F between the dark colors and the light colors is defined; The color intensity value variance F is used to distinguish between dark and light colors, and the portion where the mean of the intensity values ​​of the three color channels RGB in the first face color image or the second face color image is greater than the color intensity value variance F is a light area, and the portion where the mean of the intensity values ​​of the three color channels RGB in the first face color image or the second face color image is less than the color intensity value variance F is a dark area. The color gamut feature value P of the light area L And the dark color gamut feature value P d The input feature P is combined and the vector of P is represented as P(P L , P d ).

10. A TCM facial segmentation system based on big data as claimed in claim 9, characterized in that: The modular facial inspection segmentation model is constructed using a deep learning method that uses an improved convolutional neural network.

Images