An acupoint intelligent identification method based on augmented reality and a wearable device

Abstract

A method and wearable device for intelligent acupoint recognition based on augmented reality are disclosed. The method includes: Step 1, using a wearable device to collect RGB images, depth and distance information, and user head posture data of a target area on the user's body in real time; Step 2, training an acupoint recognition model based on YOLO, inputting the RGB images into the acupoint recognition model to identify local features of the acupoints; converting the depth and distance information into depth point cloud data, and constructing a personalized three-dimensional human body model using a Poisson surface reconstruction algorithm after voxel downsampling, and extracting key skeletal points; Step 3, mapping a pre-stored acupoint database onto the personalized three-dimensional human body model, and obtaining precise three-dimensional coordinates of the acupoints by combining the local features of the acupoints; Step 4, calculating the two-dimensional coordinates of the acupoints in the field of view through perspective projection based on the precise three-dimensional coordinates of the acupoints and the user's head posture data, and accurately overlaying and registering the acupoint names and acupoint locations onto the real body parts observed by the user.

Description

Technical Field

[0001] This invention relates to the interdisciplinary fields of traditional Chinese medicine medical devices and artificial intelligence, acupoint big data, and computer vision, and in particular to an acupoint intelligent recognition method and wearable device based on augmented reality. Background Technology

[0002] Currently, in traditional Chinese medicine acupuncture and massage therapy and health care, the location of acupoints mainly relies on the physician's experience and techniques such as "bone measurement method" and "anatomical landmark method," which has the following significant problems and defects: 1. High dependence on experience: Training a TCM doctor who can accurately locate acupoints is time-consuming and costly, and beginners and enthusiasts find it difficult to master quickly.

[0003] 2. Inconsistent standards and difficulty in controlling precision: Different physicians have different understandings and tactile sensations of anatomical landmarks, which leads to subjective deviations in the location of acupoints and affects the repeatability of therapeutic effects.

[0004] 3. Insufficient existing auxiliary tools: There are some electronic acupoint charts or simple apps on the market, but they are usually based on static two-dimensional images and cannot be matched and calibrated in real time and three-dimensional with the user's actual body parts, resulting in poor practicality.

[0005] 4. Lack of immersive guidance: Traditional methods cannot "project" the abstract acupoint theory onto the real human body, and learners and practitioners cannot obtain the intuitive experience of "what you see is what you get".

[0006] While existing technologies incorporate depth cameras, their solutions primarily focus on background computation and display, failing to provide real-time, immersive first-person perspective guidance. This invention addresses these shortcomings by offering a real-time, accurate, intuitive, and easy-to-use acupoint recognition solution. Summary of the Invention

[0007] The purpose of this invention is to address the problems of existing technologies where acupoint location relies on experience, has low accuracy, and lacks intuitive visualization guidance. It provides an augmented reality-based intelligent acupoint recognition method that accurately and in real-time overlays virtual acupoint information onto the user's real field of vision.

[0008] The present invention also provides a wearable device.

[0009] The technical solution adopted to achieve the purpose of this invention is: An augmented reality-based intelligent acupoint recognition method includes the following steps: Step 1: Use wearable devices to collect RGB images, depth and distance information, and user head posture data of the target area of ​​the user's body in real time; Step 2: Based on YOLO training, an acupoint recognition model is obtained. The RGB image from Step 1 is input into the acupoint recognition model to identify the local features of the acupoints. The depth distance information is converted into depth point cloud data. After voxel downsampling of the depth point cloud data, a personalized three-dimensional human body model is constructed using the Poisson surface reconstruction algorithm, and key points of the skeleton are extracted. Step 3: Map the pre-stored acupoint database onto the personalized 3D human body model obtained in Step 2, calculate the 3D coordinate range of each acupoint, and combine the local features of the acupoints obtained in Step 2 to obtain the precise 3D coordinates of the acupoints. Step 4: Based on the precise three-dimensional coordinates of the acupoints obtained in Step 3 and the user's head posture data in Step 1, the two-dimensional coordinates of the acupoints in the field of view are calculated through perspective projection. The graphics engine renders virtual markers and information in real time, accurately overlaying and registering the acupoint names and acupoint location information onto the real body parts observed by the user.

[0010] In the above technical solution, in step 2, the acupoint recognition model is built based on the YOLO network framework. YOLO adopts a network structure optimized for keypoint detection, including a backbone component, a neck component, and a detection head component. The three components are cascaded sequentially to form a forward inference path. The input to the backbone component is the RGB image acquired in step 1. It contains a multi-scale convolution module, a deformable convolution module, and an enhanced feature extraction module. By performing layer-by-layer feature extraction on the input RGB image, it outputs a set of basic feature maps at different scales. The input of the neck component is the multi-scale basic feature map output by the backbone component. Internally, it adopts a combination structure of feature pyramid network FPN and path aggregation network PAN to perform top-down semantic enhancement and bottom-up localization information transmission on the input multi-scale feature map, realizing deep fusion of high-resolution features and low-resolution semantic information, and outputting a set of multi-scale feature maps after fusion and enhancement. The input to the detection head component is the fused and enhanced multi-scale feature map output by the neck component. It has a key point feature channel inside, and decodes the fused and enhanced multi-scale feature map layer by layer to output the local features of the acupoints, which include the two-dimensional position coordinates of multiple acupoint key points in the image coordinate system and their visibility state prediction results.

[0011] In the above technical solution, step 2, the extraction of the skeletal key points specifically involves: performing voxel downsampling on the depth point cloud data to obtain a simplified point cloud. After downsampling, statistical filtering is used to remove samples whose distance from the neighborhood mean exceeds the neighborhood boundary. μ+2σ The outliers, among which, μThe neighborhood mean σ Using the neighborhood standard deviation, after removing outliers, the gradient field of the indicator function is obtained by reconstructing the sparse linear system using the implicit Poisson surface. A personalized 3D human body model is constructed, and the skeletal key points in the personalized 3D human body model are extracted by a skeletal key point detection algorithm based on geometric features.

[0012] In the above technical solution, the The calculation formula is: ; In the formula, For a single data point, This is the original point cloud collection. For the first i The coordinate components of each point Voxel size; The process of solving the gradient field of the indicator function involves constructing and solving the sparse linear system, and determining the gradient field of the indicator function. The calculation formula is: ; In the formula, For the first j One sampling point, For point The normal vector at that point, For the indicator function in gradient at; The formula for calculating the key points of the skeleton is as follows: ; In the formula, For local point cloud features, For lightweight keypoint detection networks, This is the key point.

[0013] In the above technical solution, in step 3, the pre-stored acupoint database is mapped onto the personalized three-dimensional human body model obtained in step 2 using an affine transformation or nonlinear registration algorithm. The affine transformation algorithm is expressed as follows: ; In the formula, Let be the affine transformation function. For rotation matrix, The anisotropic scaling matrix, Input coordinates, It is a translation vector.

[0014] A wearable device for implementing the augmented reality-based acupoint intelligent recognition method includes at least one image acquisition module for acquiring RGB images of a target area of ​​a user's body, a depth sensing module for acquiring depth and distance information of the target area of ​​the user's body, a posture sensing module for tracking the user's head movement in real time to acquire head posture data, an AR display module for projecting a personalized 3D model of mapped acupoints into the user's field of vision, a data processing module for performing steps 2 to 4, an acupoint database for storing the location and attribute information of standard acupoints, and a data storage module for storing a trained acupoint recognition model, a 3D human body model with acupoint location annotations, and other data necessary for system operation.

[0015] In the above technical solution, the smart wearable device is smart glasses; the image acquisition module is a camera; the depth perception module is a structured light, ToF, or binocular depth camera; the attitude sensing module is an IMU; the AR display module is an optical waveguide or a miniature OLED display, BirdBath, or a freeform prism; the data processing module uses a cloud server, and the device receives the results through a wireless communication module.

[0016] An electronic device includes: one or more processors; and a memory for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the augmented reality-based acupoint intelligent recognition method.

[0017] A computer-readable storage medium storing computer-executable instructions, which, when executed, are used to implement the augmented reality-based acupoint intelligent recognition method.

[0018] A computer program product includes computer-executable instructions that, when executed, implement the augmented reality-based acupoint intelligent recognition method.

[0019] Compared with the prior art, the beneficial effects of the present invention are: 1. High-precision positioning: Through multimodal data fusion of "visual recognition + depth perception + standard human acupoint database", standard acupoint data is dynamically adapted to individualized physiological structures.

[0020] 2. Real-time and intuitive: It achieves millisecond-level real-time calculation and display, so users do not need to look away from other screens during operation, realizing a seamless "first-person perspective" operation experience, improving the smoothness of operation and concentration.

[0021] 3. Lowering the professional threshold: It enables beginners and enthusiasts of traditional Chinese medicine to locate acupoints as accurately as experts, which greatly promotes the popularization and standardization of traditional Chinese medicine knowledge.

[0022] 4. Strong functional expandability: This invention can be easily expanded to include more functions, such as displaying meridian pathways, recommending acupoint combinations based on syndrome differentiation, and recording treatment history, forming a complete intelligent auxiliary ecosystem for traditional Chinese medicine.

[0023] 5. This invention is the first to achieve dynamic fusion of a standardized acupoint database and a personalized 3D human body model. Through affine transformation or nonlinear registration techniques, acupoint information based on a standard human body is intelligently adapted to user models with different body postures, thereby establishing a 3D coordinate range of acupoints with individual differences. Furthermore, this invention integrates multimodal information (3D geometric structure and RGB texture features), preserving the structural constraints of human morphology while introducing local features at the image level for fine calibration. This significantly improves the accuracy and robustness of acupoint positioning, overcoming the positioning deviations caused by human body differences in traditional methods, and providing a reliable and adaptive technical foundation for applications such as intelligent acupuncture and rehabilitation medicine. Attached Figure Description

[0024] Figure 1 The diagram shows a schematic of the acupoint intelligent recognition method based on augmented reality according to the present invention.

[0025] Figure 2 The diagram shown illustrates the principle of the human body 3D modeling and acupoint mapping algorithm of this invention.

[0026] Figure 3 The image shown is a visual effect diagram of the AR virtual-real fusion display of the present invention.

[0027] Figure 4 The diagram shows the acupoint information registered to the user's actual body parts according to the present invention. Detailed Implementation

[0028] The present invention will be further described in detail below with reference to specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0029] Example 1 Reference Figure 1 An augmented reality-based intelligent acupoint recognition method includes the following steps: Step 1, Image and Depth Information Acquisition: Real-time acquisition of RGB images, depth distance information, and user head posture data of the target area of ​​the user's body through wearable devices.

[0030] Step 2: Input the RGB image collected in Step 1 into the trained acupoint recognition model to identify acupoint features. At the same time, convert the depth distance information data collected in Step 1 into depth point cloud data. After voxel downsampling of the depth point cloud data, generate a 3D point cloud of the human target area through the Poisson surface reconstruction algorithm. Based on the 3D point cloud of the target area, construct a personalized 3D human body model in real time and extract key skeletal points.

[0031] Furthermore, the acupoint recognition model is constructed as follows: Based on the computer vision algorithm YOLO, during the data preprocessing stage, the spatial coordinates, visibility status, and corresponding local human body regions of the acupoints are uniformly converted into a standardized label format that meets the input requirements of YOLO. This enables the model to simultaneously learn the positioning information of the target region (local human body) and the precise location of the key points of the acupuncture points. YOLO is then used to train the model on the standardized label format.

[0032] The acupoint recognition model is built on the YOLO network framework. YOLO adopts a network structure optimized for keypoint detection, including a backbone component, a neck component, and a detection head component. These three components are cascaded to form a forward inference path. The backbone component takes the RGB image acquired in step 1 as input and contains a multi-scale convolution module, a deformable convolution module, and an enhanced feature extraction module. By extracting features layer by layer from the input RGB image, it outputs a set of basic feature maps at different scales. These basic feature maps contain both high-resolution shallow texture details (such as human surface texture information, skin brightness gradient changes, and local morphological features around acupoints) and low-resolution deep semantic information. The stacking of the multi-scale convolution module and the deformable convolution module can effectively capture human surface texture information, skin brightness gradient changes, and local morphological features around acupoints. In acupoint areas close to the body surface, texture changes are often weak, and their location usually depends on bony landmarks and muscle direction. The introduced enhanced feature extraction module enables the model to more robustly identify these weak but important structural information and improves the model's spatial resolution capability in fine-grained local areas of the human body.

[0033] The input to the neck component is the multi-scale basic feature map output by the backbone component. Internally, it adopts a combination structure of Feature Pyramid Network (FPN) and Path Aggregation Network (PAN) to perform top-down semantic enhancement and bottom-up localization information transfer on the input multi-scale feature map, achieving deep fusion of high-resolution features and low-resolution semantic information. The output of the neck component is a set of multi-scale feature maps after fusion and enhancement. The combination structure of Feature Pyramid Network (FPN) and Path Aggregation Network (PAN) enables the model to maintain high sensitivity to subtle texture changes and local morphological features around acupoints while perceiving a large range of human anatomical structures. This provides the detection head component with hierarchical and complete feature input, thereby improving the accuracy and robustness of acupoint key point localization.

[0034] The input to the detection head component is the fused and enhanced multi-scale feature map output by the neck component. It has a key point feature channel, which enables the model to further accurately predict the position and visibility status of multiple acupoint key points after completing the target detection task (i.e., locating local areas of the human body). The input feature map is decoded layer by layer, so that each key point can obtain the corresponding structural information on a specific feature layer, thereby enhancing the stability of the overall localization. Finally, the output is the acupoint local feature containing the two-dimensional position coordinates of multiple acupoint key points in the image coordinate system and their visibility status prediction results.

[0035] Furthermore, the extraction of the skeletal key points specifically involves: performing voxel downsampling on the depth point cloud data to obtain a simplified point cloud. After downsampling, statistical filtering is used to remove samples whose distance from the neighborhood mean exceeds the neighborhood boundary. μ+2σ Outliers are identified, where μ is the neighborhood mean and σ is the neighborhood standard deviation. After removing outliers, the gradient field of the indicator function is obtained by reconstructing the sparse linear system using the implicit Poisson surface. A personalized 3D human body model is constructed, and a skeletal key point detection algorithm based on geometric features is used to extract the skeletal key points in the personalized 3D human body model.

[0036] Furthermore, the simplified point cloud The calculation formula is: ; In the formula, For a single data point, This is the original point cloud collection. For the first i The coordinate components of each point Voxel size.

[0037] The formula for calculating the gradient field of the indicator function is as follows: ; In the formula, For the first j One sampling point, For point The normal vector at that point, For the indicator function in The gradient at that point.

[0038] The extraction of elbow and wrist joints includes 18 key points, and the detection formula for these key points is as follows: ; In the formula, For local point cloud features, For lightweight keypoint detection networks, As a key point, in this embodiment, .

[0039] Step 3: The pre-stored acupoint database with acupoint location annotations is dynamically adapted to the personalized 3D human body model generated in Step 2 using affine transformation or a nonlinear registration algorithm. The 3D coordinate range of each acupoint on the current user's body is calculated. Combining the acupoint features identified in Step 2 with the 3D coordinate range of each acupoint on the previous user's body, the acupoints are accurately located, yielding precise 3D coordinates. The principle of the human 3D modeling and acupoint mapping algorithm is explained in [reference needed]. Figure 2 , refer to Figure 3 This is a visual effect diagram of AR virtual-real fusion display.

[0040] Furthermore, in this embodiment, an affine transformation registration algorithm is used to calculate the three-dimensional coordinate range of each acupoint on the current user's body. The formula for the transformation registration algorithm is: ; In the formula, Let be the affine transformation function. For rotation matrix (orthogonal matrix, ), This is an anisotropic scale matrix (reflecting the differences in the length-to-width ratio of individual limbs). , , They are respectively axis, shaft and Axial scaling factor of the axis. Input coordinates (i.e., points in the pre-stored acupoint database). It is a translation vector.

[0041] Step 4: Based on the precise 3D coordinates of the acupoints obtained in Step 3 and the user's head posture data collected in Step 1, a visual algorithm is used to register the personalized 3D model of the mapped acupoints from Step 3 with the real-time image. This calculates the precise pose of the actual body part relative to the wearable device, thus determining the virtual-real spatial transformation relationship. Then, based on this transformation relationship, the 2D coordinates of the acupoints in the field of view are calculated using perspective projection. The graphics engine then renders virtual markers and information in real time, accurately overlaying and registering the acupoint names and location information onto the actual body part observed by the user. Finally, the generated virtual image is fused with real light on the user's retina, referencing... Figure 4 This creates a visual effect where information is stably "fitted" onto the body's surface.

[0042] Example 2 A wearable device for implementing the augmented reality-based acupoint intelligent recognition method in Embodiment 1 includes: at least one image acquisition module for acquiring RGB images of a target area of ​​a user's body, a depth perception module for acquiring depth information of the target area of ​​the user's body, a posture sensing module for tracking the user's head movement in real time to acquire head posture data, an AR display module for projecting a personalized 3D model of mapped acupoints into the user's field of vision, a data processing module for executing steps 2 to 4 of Embodiment 1, an acupoint database for storing the location and attribute information of standard acupoints, and a data storage module for storing a trained acupoint recognition model, a 3D human body model with acupoint location annotations, and other data necessary for system operation.

[0043] Furthermore, in this embodiment, the smart wearable device is smart glasses; the image acquisition module is a camera; the depth perception module is a structured light, ToF, or binocular depth camera; the attitude sensing module is an IMU; the AR display module is an optical waveguide or a miniature OLED display, BirdBath, or a freeform prism; the data processing module can also be a cloud server, and the device receives the results through a wireless communication module to reduce the power consumption and computational load of the local device.

[0044] Example 3 This embodiment describes a specific method of using the wearable device in Embodiment 2, including the following steps: Step 1: The user puts on the smart glasses and places their arm in front of their field of vision.

[0045] Step 2: Start the device. The camera and depth sensor will begin to work, acquiring RGB images and point cloud data of the arm.

[0046] Step 3: The data processing module identifies key points such as the elbow and wrist joints of the user's arm and constructs a three-dimensional mesh model of the arm.

[0047] Step 4: The system retrieves the standard location of the "Quchi" acupoint (located at the outer end of the elbow crease) from the acupoint database, maps it onto the user's personalized arm model using an algorithm, and calculates its precise three-dimensional location on the current user.

[0048] Step 5: Based on the 3D position calculated in Step 3 and the real-time posture of the user's head, the AR display module renders a flashing dot mark near the elbow of the user's real arm in the user's field of vision, and displays information such as "Quchi LI11" in a text box next to it.

[0049] Step 6: The user performs subsequent pressing or acupuncture operations based on the AR markers in Step 5.

[0050] Example 4 This embodiment provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the augmented reality-based acupoint intelligent recognition method described in Embodiment 1 above.

[0051] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of hardware embodiments, software embodiments, or embodiments combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.

[0052] Example 5 This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in the augmented reality-based acupoint intelligent recognition method described in Embodiment 1 above.

[0053] The computer-readable storage medium in this embodiment can be an internal storage unit of the terminal, such as the terminal's hard disk or memory; the computer-readable storage medium in this embodiment can also be an external storage device of the terminal, such as a plug-in hard disk, smart memory card, secure digital card, flash memory card, etc. equipped on the terminal; furthermore, the computer-readable storage medium can include both the terminal's internal storage unit and external storage devices.

[0054] The computer-readable storage medium of this embodiment is used to store computer programs and other programs and data required by the terminal. The computer-readable storage medium can also be used to temporarily store data that has been output or will be output.

[0055] Example 6 This embodiment provides a computer program product including computer-executable instructions stored on a machine-readable storage medium (such as a disk, flash memory, or optical disk). When executed by at least one data processing device (such as a microprocessor or digital signal processor), the instructions cause the data processing device to perform the steps in the augmented reality-based acupoint intelligent recognition method described in Embodiment 1.

[0056] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0057] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0058] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0059] The above description is only a preferred embodiment of the present invention. It should be noted that, for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for intelligent acupoint recognition based on augmented reality, characterized in that, Includes the following steps: Step 1: Use wearable devices to collect RGB images, depth and distance information, and user head posture data of the target area of ​​the user's body in real time; Step 2: Train an acupoint recognition model based on YOLO, and input the RGB image from Step 1 into the acupoint recognition model to identify the local features of the acupoints; Depth distance information is converted into depth point cloud data. After voxel downsampling of the depth point cloud data, a personalized 3D human body model is constructed using the Poisson surface reconstruction algorithm, and key points of the skeleton are extracted. Step 3: Map the pre-stored acupoint database onto the personalized 3D human body model obtained in Step 2, calculate the 3D coordinate range of each acupoint, and combine the local features of the acupoints obtained in Step 2 to obtain the precise 3D coordinates of the acupoints. Step 4: Based on the precise three-dimensional coordinates of the acupoints obtained in Step 3 and the user's head posture data in Step 1, the two-dimensional coordinates of the acupoints in the field of view are calculated through perspective projection. The graphics engine renders virtual markers and information in real time, accurately overlaying and registering the acupoint names and acupoint location information onto the real body parts observed by the user.

2. The acupoint intelligent recognition method based on augmented reality as described in claim 1, characterized in that, In step 2, the acupoint recognition model is built based on the YOLO network framework. YOLO adopts a network structure optimized for keypoint detection, including a backbone component, a neck component, and a detection head component. The three components are cascaded to form a forward inference path. The input to the backbone component is the RGB image acquired in step 1. It contains a multi-scale convolution module, a deformable convolution module, and an enhanced feature extraction module. By performing layer-by-layer feature extraction on the input RGB image, it outputs a set of basic feature maps at different scales. The input of the neck component is the multi-scale basic feature map output by the backbone component. Internally, it adopts a combination structure of feature pyramid network FPN and path aggregation network PAN to perform top-down semantic enhancement and bottom-up localization information transmission on the input multi-scale feature map, realizing deep fusion of high-resolution features and low-resolution semantic information, and outputting a set of multi-scale feature maps after fusion and enhancement. The input to the detection head component is the fused and enhanced multi-scale feature map output by the neck component. It has a key point feature channel inside, and decodes the fused and enhanced multi-scale feature map layer by layer to output the local features of the acupoints, which include the two-dimensional position coordinates of multiple acupoint key points in the image coordinate system and their visibility state prediction results.

3. The acupoint intelligent recognition method based on augmented reality as described in claim 1, characterized in that, In step 2, the extraction of the skeletal key points specifically involves: performing voxel downsampling on the depth point cloud data to obtain a simplified point cloud. After downsampling, statistical filtering is used to remove samples whose distance from the neighborhood mean exceeds the neighborhood boundary. μ+2σ The outliers, among which, μ The neighborhood mean σ Using the neighborhood standard deviation, after removing outliers, the gradient field of the indicator function is obtained by reconstructing the sparse linear system using the implicit Poisson surface. A personalized 3D human body model is constructed, and the skeletal key points in the personalized 3D human body model are extracted by a skeletal key point detection algorithm based on geometric features.

4. The acupoint intelligent recognition method based on augmented reality as described in claim 3, characterized in that, The The calculation formula is: ； In the formula, For a single data point, This is the original point cloud collection. For the first i The coordinate components of each point Voxel size; The process of solving the gradient field of the indicator function involves constructing and solving the sparse linear system, and determining the gradient field of the indicator function. The calculation formula is: ； In the formula, For the first j One sampling point, For point The normal vector at that point, For the indicator function in gradient at; The formula for calculating the key points of the skeleton is as follows: ； In the formula, For local point cloud features, For lightweight keypoint detection networks, This is the key point.

5. The acupoint intelligent recognition method based on augmented reality as described in claim 1, characterized in that, In step 3, the pre-stored acupoint database is mapped onto the personalized 3D human body model obtained in step 2 using an affine transformation or nonlinear registration algorithm. The affine transformation algorithm is expressed as follows: ； In the formula, Let be the affine transformation function. For rotation matrix, The anisotropic scaling matrix, Input coordinates, It is a translation vector.

6. A wearable device that implements the augmented reality-based acupoint intelligent recognition method as described in claim 1, characterized in that, It includes at least one image acquisition module for acquiring RGB images of the target area of ​​the user's body, a depth perception module for acquiring depth and distance information of the target area of ​​the user's body, a posture sensing module for tracking the user's head movement in real time to acquire head posture data, an AR display module for projecting a personalized 3D model of mapped acupoints into the user's field of vision, a data processing module for performing steps 2 to 4, an acupoint database for storing the location and attribute information of standard acupoints, and a data storage module for storing the trained acupoint recognition model, a 3D human body model with acupoint location annotations, and other data necessary for system operation.

7. The wearable device as described in claim 6, characterized in that, The smart wearable device is smart glasses; the image acquisition module is a camera; the depth sensing module is a structured light, ToF, or binocular depth camera; the attitude sensing module is an IMU; the AR display module is an optical waveguide or a miniature OLED display, BirdBath, or a freeform prism; the data processing module uses a cloud server, and the device receives the results through a wireless communication module.

8. An electronic device, characterized in that, include: One or more processors; A memory for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the augmented reality-based acupoint intelligent recognition method as described in claim 1.

9. A computer-readable storage medium, characterized in that, It stores computer-executable instructions, which, when executed, are used to implement the augmented reality-based acupoint intelligent recognition method as described in claim 1.

10. A computer program product, characterized in that, The aforementioned computer program product includes computer-executable instructions, which, when executed, are used to implement the augmented reality-based acupoint intelligent recognition method as described in claim 1.

Images