Acupuncture point positioning method and system based on machine learning

Abstract

The invention belongs to the technical field of machine learning, and provides an acupoint positioning method and system based on machine learning, and the method comprises the steps: scanning a human body image, and obtaining human body data; preprocessing the human body data to obtain feature data; performing machine learning of multiple models based on the feature data; the acupoints are positioned through the multiple models obtained after machine learning; and displaying the positioned acupuncture points, and evaluating the positioning precision of the acupuncture points. The human body is scanned according to the big data of the acupoint knowledge graph during acupoint positioning so as to acquire the human body data, then the multiple models are constructed by processing and learning the human body data, finally the acupoints are positioned through the multiple models, and the traditional Chinese medicine acupoint knowledge graph is optimized according to the positioning structure, so that the accuracy of acupoint positioning is improved. In the process, the acupoint distribution rule and the acupoint reference coordinates of big data of the traditional Chinese medicine acupoint mapping knowledge domain are used for analyzing the data and accurately positioning the acupoints, which is greatly helpful to research on the acupoints of the human body.

Description

Technical Field

[0001] The present invention belongs to the technical field of machine learning, and in particular relates to an acupoint positioning method and system based on machine learning. Background Art

[0002] Existing TCM acupuncture point selection basically relies on traditional clinical experience and touch methods. Due to the differences in people's ages, heights, and body shapes, there are also great differences in acupuncture points on the human body. Inaccurate acupuncture point selection leads to ineffective treatment effects of acupuncture and moxibustion. Traditional accurate acupuncture point selection requires experienced TCM practitioners and clinical experience, which greatly consumes TCM resources. In the early stages of learning the locations of human acupuncture points, there is no accurate acupuncture point reference, which can easily lead to ineffective acupuncture and moxibustion effects, and even the possibility of misdiagnosis. Summary of the invention

[0003] In order to solve the problems in the background technology, the present invention proposes an acupoint positioning method and system based on machine learning.

[0004] In order to achieve the above object, the present invention adopts the following technical solutions: A method for locating acupoints based on machine learning, comprising the following steps: Based on the acupoint knowledge graph, scan human body images and obtain human body data; Preprocessing human body data to obtain characteristic data related to acupoints in the human body data; Performing machine learning of multiple models based on the feature data, wherein the multiple models include a human acupoint model, an acupoint association model, and a human acupoint positioning standard value; Acupuncture points are located through multiple models after machine learning; The located acupoints are displayed and the positioning accuracy of the acupoints is evaluated.

[0005] Preferably, scanning a human body image based on an acupoint knowledge graph to obtain human body data comprises the following steps: Obtain the three-dimensional coordinate data of acupuncture points on the human body through imaging equipment; Use imaging equipment to collect human motion dimension data through voice guidance; Human height and finger width data are collected through imaging equipment.

[0006] Preferably, preprocessing the human body data to obtain characteristic data related to acupoints in the human body data includes the following steps: Integrate, screen, classify, denoise, filter and normalize human body data; After normalization, the shape, size, dimension and position of the acupoints are extracted in combination with the knowledge graph of TCM acupoints.

[0007] Preferably, performing machine learning of multiple models based on feature data comprises the following steps: Establish a human acupoint model based on three-dimensional coordinate data and acupoint knowledge graph; Establishing local coordinates of the acupoint area based on the human body motion size data, and establishing an acupoint association model through the local coordinates; the acupoint association model is associated with an acupoint knowledge graph, and / or a coordinate graph of a distorted area of ​​the acupoint after the human body motion; Parameters for calculating human acupoints are collected based on human height and finger width, and standard values ​​for positioning human acupoints are established through the parameters for calculating human acupoints.

[0008] Preferably, during the machine learning process: The weight parameters of the human acupoint model are updated through the back-propagation algorithm and optimizer to minimize the error between the predicted position and the actual position; The mean square error loss function is used to measure the accuracy of acupoints.

[0009] Preferably, locating acupoints by using a multiple model after machine learning comprises the following steps: Determine the human body acupoint positioning standard values ​​corresponding to the height and finger width of the human body to be acupoint positioned; Based on the human acupoint positioning standard value, the acupoint data to be positioned is input into the human acupoint model to determine the coordinates of the human acupoint, and / or the acupoint data to be positioned is input into the acupoint association model to determine the acupoint association coordinates.

[0010] Preferably, displaying the located acupoints and evaluating the positioning accuracy of the acupoints comprises the following steps: Perform moxibustion or acupuncture verification based on the coordinates of human acupoints and acupoint-related coordinates to determine whether the acupoint positioning is accurate; If it is accurate, the acupoint knowledge graph is updated; If inaccurate, retrain the multiple models.

[0011] An acupoint positioning system based on machine learning, comprising: A scanning unit, used to scan human body images and obtain human body data based on acupoint knowledge graph; A preprocessing unit, used for preprocessing human body data to obtain characteristic data related to acupoints in the human body data; A training unit, used for performing machine learning of multiple models based on feature data, wherein the multiple models include a human acupoint model, an acupoint association model, and a human acupoint positioning standard value; A positioning unit, used to locate acupuncture points through multiple models after machine learning; The optimization unit is used to display the located acupoints and evaluate the positioning accuracy of the acupoints.

[0012] Preferably, the scanning unit comprises: A first scanning module is used to obtain three-dimensional coordinate data of acupuncture points of the human body through an imaging device; The second scanning module is used to collect human body motion size data using an imaging device under voice guidance; The third scanning module is used to collect human height and finger width data through imaging equipment.

[0013] Preferably, the training unit comprises: The first training module is used to establish a human acupoint model based on three-dimensional coordinate data and acupoint knowledge graph; The second training module is used to establish local coordinates of the acupoint area based on the human body motion size data, and to establish an acupoint association model through the local coordinates; the acupoint association model is associated with an acupoint knowledge graph, and / or a coordinate graph of a distorted area of ​​the acupoint after the human body motion is associated; The third training module is used to collect parameters for calculating human acupoints based on human height and finger width, and to establish standard values ​​for positioning human acupoints through the parameters for calculating human acupoints.

[0014] Preferably, the positioning unit comprises: The first positioning module is used to determine the human body acupoint positioning standard values ​​corresponding to the height and finger width of the human body to be acupoint positioned; The second positioning module is used to input the acupoint data to be located into the human acupoint model to determine the coordinates of the human acupoint based on the human acupoint positioning standard value, and / or input the acupoint data to be located into the acupoint association model to determine the acupoint association coordinates.

[0015] Preferably, the optimization unit comprises: A verification module is used to verify moxibustion or acupuncture based on the coordinates of human acupoints and the coordinates associated with the acupoints to determine whether the acupoint positioning is accurate; Optimization module, used to update the acupoint knowledge graph when the acupoints are accurately located; The return module is used to retrain multiple models when acupoint positioning is inaccurate.

[0016] Beneficial effects of the present invention: When locating acupoints, the present invention scans the human body according to the acupoint knowledge graph big data, aiming to obtain human body data, and then constructs multiple models by processing and learning the human body data. Finally, the acupoints are located through multiple modules, and the TCM acupoint knowledge graph is optimized according to the positioning structure. This process uses the acupoint distribution rules and acupoint reference coordinates of the TCM acupoint knowledge graph big data to analyze the data and accurately locate the acupoints, which is of great help to the research on human acupoints.

[0017] Other features and advantages of the present invention will be described in the following description, and partly become obvious from the description, or be understood by implementing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures pointed out in the description and the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0019] Figure 1 A general flow chart of an acupoint location method based on machine learning of the present invention is shown; Figure 2 The acupoint knowledge graph of the present invention is shown; Figure 3 A coordinate diagram of a scanned human body image of the present invention is shown; Figure 4 A detailed flow chart of acupoint positioning of the present invention is shown; Figure 5 A schematic diagram of the acupoint association model of the present invention is shown. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] A method for locating acupoints based on machine learning, such as Figure 1 As shown, the following steps are included: S1: Based on the acupoint knowledge graph, scan the human body image and obtain human body data.

[0022] S2: Preprocess the human body data to obtain feature data related to acupoints in the human body data.

[0023] S3: Performing machine learning of multiple models based on the feature data, wherein the multiple models include a human acupoint model, an acupoint association model, and a human acupoint positioning standard value.

[0024] S4: Acupoints are located through multiple models after machine learning.

[0025] S5: Display the located acupoints and evaluate the positioning accuracy of the acupoints.

[0026] It should be noted that in Figure 1 In the system, acupoint positioning adopts artificial intelligence algorithms and TCM acupoint knowledge graph big data to help accurately determine the location of human acupoints; the system uses artificial intelligence technologies such as sensors, image recognition and machine learning, collects acupoint measurement standards, scans the human body surface and related movements, and uses the acupoint distribution rules and acupoint reference coordinates of TCM acupoint knowledge graph big data to analyze data and accurately locate acupoints, which is of great help in the research of human acupoints.

[0027] Exemplarily, in S1: scanning a human body image based on an acupoint knowledge graph to obtain human body data includes the following steps: S101: Acquire the three-dimensional coordinate data of acupuncture points of the human body through imaging equipment; for example, use a 3D scanner to perform a full-body scan on the subject to obtain the three-dimensional coordinate data of the acupuncture points. S102: Use imaging equipment to collect human motion size data through voice guidance; for example, use a voice assistant to guide the subject to perform specific movements, such as bending over, raising hands, etc., and use a motion capture device to record the motion size data. S103: Collect human height and finger width data through imaging equipment; for example, use a height measuring instrument to obtain the height of the subject, and use a finger width measuring tool to obtain finger width data.

[0028] It should be noted that in S101, the human body can be allowed to stand, lie flat or lie prone. At this time, the coordinates of the human acupuncture points can be collected by a scanner. The purpose of this step is to obtain the coordinates of the acupuncture points of the human body in a relaxed state without movement. The initial positions of the acupuncture point coordinates can be combined with Figure 2 The knowledge graph of TCM acupoints. In S102, after the human body's movements are guided by voice, certain distortions will occur in the areas corresponding to the acupoints. Therefore, S102 can collect all the associated coordinates at the distorted locations to provide data support for the subsequent acupoint positioning of the human body in motion. In S103, the main consideration is the differences in body shape between humans. Generally speaking, there are slight differences in the acupoint coordinates of each individual. Therefore, collecting data such as height and finger width is conducive to summarizing the changes in acupoints between individuals with different differences during the machine learning process.

[0029] It needs to be further explained that if Figure 3 As shown, the following coordinate systems are involved when scanning a human body: O w -X w Y wZ w : World coordinate system, describing the camera position, unit m; Oc-XcYcZc : Camera coordinate system, with the optical center as the origin, unit m; o-xy : Image coordinate system, the origin is the center of the imaging plane, unit is mm; P : A point in the world coordinate system; p :point P The imaging point in the image has the following coordinates in the image coordinate system: x,y ), the coordinates in the pixel coordinate system are ( u,v ); uv : Pixel coordinate point, the origin is the upper left corner of the image, the unit is pixel (Pixel), the coordinate of each pixel ( u,v ) are the column and row numbers of the pixel in the array respectively; f : Camera focal length, equal to o to Oc distance.

[0030] Exemplarily, in S2, preprocessing the human body data to obtain characteristic data related to acupoints in the human body data includes the following steps: S201: Integrate the data collected in S1.

[0031] S202: Screening valid data: Use algorithms and data mining techniques to clean and screen the collected data. Eliminate duplicate data entries, identify and eliminate erroneous or inconsistent information, and ensure the accuracy and reliability of the data.

[0032] S203: Classification processing: Classify the data according to the meridians, functions, locations, etc. of the acupoints. For example, the acupoints can be divided into twelve regular meridian acupoints, eight extraordinary meridian acupoints, etc. according to the meridians; the acupoints can be divided into acupoints for treating headaches, acupoints for treating insomnia, etc. according to their functions.

[0033] S204: Data denoising: There may be some outliers or erroneous data in the data set, which will interfere with subsequent analysis and processing. Therefore, it is necessary to apply denoising algorithms, such as statistical methods, machine learning methods, etc., to identify and exclude these outliers.

[0034] S205: Filtering: In order to smooth the data and eliminate random noise, a filtering algorithm can be applied. Common filtering algorithms include Gaussian filtering, mean filtering, median filtering, etc. According to the characteristics and requirements of the data, select a suitable filtering algorithm for processing.

[0035] S206: Normalization: Since the acupoint data from different sources may have different scales or ranges, the data needs to be normalized. The purpose of normalization is to standardize the data to a uniform scale or range for subsequent analysis and comparison. Normalization methods include linear normalization, nonlinear normalization, etc.

[0036] S207: Combine with the knowledge graph of TCM acupoints: In the normalized acupoint data, combine with the knowledge graph of TCM acupoints to further analyze the attributes of the acupoints. The knowledge graph of TCM acupoints contains a large amount of feature data about acupoints, such as the shape, size, size and position of the acupoints. Through reasoning and analysis of the knowledge graph, the required acupoint attribute information can be extracted.

[0037] S208: Establishing an acupoint database: Finally, the acupoint data after integration, screening, classification, denoising, filtering, and normalization, as well as the acupoint attribute information extracted from the TCM acupoint knowledge graph, are stored in the acupoint database. The acupoint database should contain detailed acupoint information to facilitate subsequent research and application.

[0038] Exemplarily, performing machine learning of multiple models based on feature data in S3 includes the following steps: S301: Establish a human acupoint model based on three-dimensional coordinate data and acupoint knowledge graph; S302: Establish local coordinates of the acupoint area based on human motion size data, and establish an acupoint association model through the local coordinates; the acupoint can be associated with the model-associated acupoint knowledge graph, and / or the coordinate graph of the distorted area of ​​the acupoint after the associated human motion; S303: Collect parameters for calculating human acupoints based on human height and finger width, and establish standard positioning values ​​for human acupoints through the parameters calculated for human acupoints.

[0039] It should be noted that, in the process of machine learning in S301, the weight parameters of the human acupoint model can be updated by the back propagation algorithm and the optimizer to minimize the error between the predicted position and the actual position. In addition, the mean square error loss function is used to measure the accuracy of the acupoints, and its calculation formula is as follows: Mean square error MSE = \frac{1}{n} \sum_{i=1}^{n} (Y_i - \hat{Y}_i)^ 2 Where (n) is the number of samples; (Y_i) is the actual value of the i-th sample; (\hat{Y}_i) is the model's predicted value for a sample; frac{1}{n} means taking the average of all error terms, where ( \frac{}{} ) is a fractional representation, representing a division operation; \sum_{i=1}^{n} : means summing all items from ( i=1 ) to ( i=n ).

[0040] Furthermore, if Figure 4 As shown, the human acupoint model needs to be combined with the existing acupoint knowledge graph, which requires scanning the entire human body, while the acupoint association model can scan the corresponding position according to the action needs, such as the height, finger width, etc. mentioned above.

[0041] In addition, natural language processing and other technologies are used when establishing the association model to associate the acupoint association model with various situations.

[0042] like Figure 5 As shown in the figure, the acupoint association model can identify entities such as disease names, acupoint names, etc. from clinical data and the original TCM acupoint knowledge graph. Then, these entities are associated through semantic matching and other technologies to establish connections between them. On the basis of entity recognition and association, knowledge reasoning and mining technology are used to discover potential connections and rules in clinical data and the original TCM acupoint knowledge graph. For example, the association between a certain disease and a specific acupoint prescription (a combination of a single or multiple acupoints), or the therapeutic effect of a certain moxibustion method on a specific disease, etc. can be mined. In addition, the acupoint association model can also associate the coordinate map of the acupoint distortion area after human body movement. For example, after the human hand moves (such as clenching a fist, opening it), the position of the acupoint will be distorted due to muscle movement. Therefore, the distorted position coordinates can be associated through the acupoint association model. Specifically, a Bayesian model can be used to facilitate the position judgment of the acupoint.

[0043] Of course, it is also possible to establish a connection between the acupoint knowledge graph and the coordinate graph of the distorted acupoint area, find the corresponding acupoint prescription through the distorted acupoint location, and finally determine the maximum weight of the acupoint. In general, the acupoint association model is a large model based on comprehensive judgment of multiple conditions. It is a data model based on convolutional neural networks. They are mainly used for image processing, but can also be used for other types of input. These convolutional layers will gradually shrink as the network deepens. Usually the number of convolutional layers is a factor of the input. In addition to the convolutional layer, there is also a feature pooling layer. Pooling is a method of filtering details: the most commonly used pooling technique is max pooling, which determines the acupoints of the human body based on conditional features.

[0044] According to the results of knowledge reasoning and mining, the relationship between clinical data and the original TCM acupoint knowledge graph is described by the Spearman correlation coefficient. The specific formula is as follows: ρ=1-[6∑[(di^2)] / [k(k^2-1)]; Where ρ is the Spearman rank correlation coefficient; di is the difference between the original value and the rank value of the i-th data point; k is the number of data points.

[0045] By calculating the Spearman correlation coefficient, we can more accurately determine whether there is a correlation between two variables. The value range of the Spearman correlation coefficient is between 0 and 1. If ρ=0, it means that there is no correlation between the two variables; if ρ=1, it means that the correlation between the two variables is very strong; if ρ=-1, it means that the opposite changes of the two variables are very strong. The Spearman correlation coefficient can be used to evaluate the relationship between categorical variables and the nonlinear relationship between quantitative variables. It is usually used to describe the relationship between a quantitative variable and a categorical variable, and can be used to find the correlation between different categories.

[0046] Exemplarily, the acupoints are located by using the multiple models after machine learning in S4, including the following steps: S401: Determine the human acupoint positioning standard values ​​corresponding to the height and finger width of the human body to be located; S402: Based on the human acupoint positioning standard values, input the acupoint data to be located into the human acupoint model to determine the coordinates of the human acupoint, and / or, input the acupoint data to be located into the acupoint association model to determine the acupoint association coordinates.

[0047] It should be noted that in S401, since each individual has differences, it is necessary to confirm the positioning standard value of the acupoint to be positioned, and then execute S402 and S403. The positioning method of S4 can accurately locate acupoints for different individuals, bringing convenience to medical personnel.

[0048] Exemplarily, displaying the located acupoints in S5 and evaluating the positioning accuracy of the acupoints includes the following steps: S501: Perform moxibustion or acupuncture verification based on the coordinates of the human acupoints and the acupoint-related coordinates to determine whether the acupoint positioning is accurate; if accurate, update the acupoint knowledge graph; if inaccurate, retrain the multiple models.

[0049] It should be noted that the update of the acupoint knowledge graph is an iterative learning process. With more iterations, the population to which the acupoint knowledge graph is applicable will gradually expand, and the location of the acupoints will be determined more accurately.

[0050] A machine learning-based acupoint positioning system includes a scanning unit, a preprocessing unit, a training unit, a positioning unit, and an optimization unit. The scanning unit is used to scan a human body image based on an acupoint knowledge graph to obtain human body data; the preprocessing unit is used to preprocess the human body data to obtain feature data related to acupoints in the human body data; the training unit is used to perform machine learning of multiple models based on the feature data, and the multiple models include a human acupoint model, an acupoint association model, and a human acupoint positioning standard value; the positioning unit is used to locate acupoints through the multiple models after machine learning; the optimization unit is used to display the located acupoints and evaluate the positioning accuracy of the acupoints.

[0051] Exemplarily, the scanning unit includes a first scanning module, a second scanning module and a third scanning module. The first scanning module is used to obtain the three-dimensional coordinate data of the acupuncture points of the human body through an imaging device; the second scanning module is used to collect the human body movement size data through the imaging device under voice guidance; and the third scanning module is used to collect the human body height and finger width data through the imaging device.

[0052] Exemplarily, the training unit includes a first training module, a second training module and a third training module. The first training module is used to establish a human acupoint model based on three-dimensional coordinate data and acupoint knowledge graph; the second training module is used to establish local coordinates of acupoint areas based on human motion size data, and establish an acupoint association model through local coordinates; the third training module is used to collect parameters for human acupoint calculation based on human height and finger width, and establish positioning standard values ​​of human acupoints through the parameters of human acupoint calculation.

[0053] Exemplarily, the positioning unit includes a first positioning module and a second positioning module. The first positioning module is used to determine the human body acupoint positioning standard value corresponding to the height and finger width of the human body to be positioned; the second positioning module is used to input the acupoint data to be positioned into the human body acupoint model to determine the coordinates of the human body acupoint based on the human body acupoint positioning standard value, and / or input the acupoint data to be positioned into the acupoint association model to determine the acupoint association coordinates.

[0054] Exemplarily, the optimization unit includes a verification module, an optimization module, and a return module. The verification module is used to verify moxibustion or acupuncture based on the coordinates of human acupoints and acupoint-related coordinates to determine whether the acupoint positioning is accurate; the optimization module is used to update the acupoint knowledge graph when the acupoint positioning is accurate; and the return module is used to retrain multiple models when the acupoint positioning is inaccurate.

[0055] Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent substitutions for some of the technical features therein; and these modifications or substitutions do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for locating acupoints based on machine learning, characterized in that: The following steps are involved: Based on the acupoint knowledge graph, scan human body images and obtain human body data; Preprocessing human body data to obtain characteristic data related to acupoints in the human body data; Performing machine learning of multiple models based on the feature data, wherein the multiple models include a human acupoint model, an acupoint association model, and a human acupoint positioning standard value; Acupuncture points are located through multiple models after machine learning; The located acupoints are displayed and the positioning accuracy of the acupoints is evaluated.

2. The acupoint locating method based on machine learning according to claim 1, characterized in that: Based on the acupoint knowledge graph, human body images are scanned to obtain human body data, including the following steps: Obtain the three-dimensional coordinate data of acupuncture points on the human body through imaging equipment; Use imaging equipment to collect human motion dimension data through voice guidance; Human height and finger width data are collected through imaging equipment.

3. The acupoint locating method based on machine learning according to claim 2, characterized in that: Preprocessing the human body data to obtain characteristic data related to acupoints in the human body data includes the following steps: Integrate, screen, classify, denoise, filter and normalize human body data; After normalization, the shape, size, dimension and position of the acupoints are extracted in combination with the knowledge graph of TCM acupoints.

4. The acupoint locating method based on machine learning according to claim 2, characterized in that: The machine learning of multiple models based on feature data includes the following steps: Establish a human acupoint model based on three-dimensional coordinate data and acupoint knowledge graph; Establishing local coordinates of the acupoint area based on the human body motion size data, and establishing an acupoint association model through the local coordinates; the acupoint association model is associated with an acupoint knowledge graph, and / or a coordinate graph of a distorted area of ​​the acupoint after the human body motion; Parameters for calculating human acupoints are collected based on human height and finger width, and standard values ​​for positioning human acupoints are established through the parameters for calculating human acupoints.

5. The acupoint locating method based on machine learning according to claim 4, characterized in that: During the machine learning process: The weight parameters of the human acupoint model are updated through the back-propagation algorithm and optimizer to minimize the error between the predicted position and the actual position; The mean square error loss function is used to measure the accuracy of acupoints.

6. The acupoint locating method based on machine learning according to claim 5, characterized in that: The acupoints are located through multiple models after machine learning, including the following steps: Determine the human body acupoint positioning standard values ​​corresponding to the height and finger width of the human body to be acupoint positioned; Based on the human acupoint positioning standard value, the acupoint data to be positioned is input into the human acupoint model to determine the coordinates of the human acupoint, and / or the acupoint data to be positioned is input into the acupoint association model to determine the acupoint association coordinates.

7. The acupoint locating method based on machine learning according to claim 6, characterized in that: Displaying the located acupoints and evaluating the positioning accuracy of the acupoints includes the following steps: Perform moxibustion or acupuncture verification based on the coordinates of human acupoints and acupoint-related coordinates to determine whether the acupoint positioning is accurate; If it is accurate, the acupoint knowledge graph is updated; If inaccurate, retrain the multiple models.

8. An acupoint positioning system based on machine learning, characterized in that: include: A scanning unit, used to scan human body images and obtain human body data based on acupoint knowledge graph; A preprocessing unit, used for preprocessing human body data to obtain characteristic data related to acupoints in the human body data; A training unit, used for performing machine learning of multiple models based on feature data, wherein the multiple models include a human acupoint model, an acupoint association model, and a human acupoint positioning standard value; A positioning unit, used to locate acupuncture points through multiple models after machine learning; The optimization unit is used to display the located acupoints and evaluate the positioning accuracy of the acupoints.

9. The acupoint positioning system based on machine learning according to claim 8, characterized in that: The scanning unit comprises: A first scanning module is used to obtain three-dimensional coordinate data of acupuncture points of the human body through an imaging device; The second scanning module is used to collect human body motion size data using an imaging device under voice guidance; The third scanning module is used to collect human height and finger width data through imaging equipment.

10. The acupoint positioning system based on machine learning according to claim 8, characterized in that: The training unit comprises: The first training module is used to establish a human acupoint model based on three-dimensional coordinate data and acupoint knowledge graph; The second training module is used to establish local coordinates of the acupoint area based on the human body motion size data, and to establish an acupoint association model through the local coordinates; the acupoint association model is associated with an acupoint knowledge graph, and / or a coordinate graph of a distorted area of ​​the acupoint after the human body motion; The third training module is used to collect parameters for calculating human acupoints based on human height and finger width, and to establish standard values ​​for positioning human acupoints through the parameters for calculating human acupoints.

11. The acupoint positioning system based on machine learning according to claim 8, characterized in that: The positioning unit comprises: The first positioning module is used to determine the human body acupoint positioning standard values ​​corresponding to the height and finger width of the human body to be acupoint positioned; The second positioning module is used to input the acupoint data to be located into the human acupoint model to determine the coordinates of the human acupoint based on the human acupoint positioning standard value, and / or input the acupoint data to be located into the acupoint association model to determine the acupoint association coordinates.

12. The acupoint positioning system based on machine learning according to claim 8, characterized in that: The optimization unit comprises: A verification module is used to verify moxibustion or acupuncture based on the coordinates of human acupoints and the coordinates associated with the acupoints to determine whether the acupoint positioning is accurate; Optimization module, used to update the acupoint knowledge graph when the acupoints are accurately located; The return module is used to retrain multiple models when acupoint positioning is inaccurate.

Images