An acupoint positioning man-machine intelligent cooperation method, device, equipment and medium

Abstract

The application discloses an acupoint positioning human-computer intelligent cooperation method and device, equipment and medium, which are applied to the technical field of computers and comprise the following steps: obtaining three-dimensional acupoint coordinates of a current object through an artificial intelligence positioning model and outputting the three-dimensional acupoint coordinates to an interactive visual interface; when an acupoint coordinate adjustment instruction is generated in the interactive visual interface, obtaining a three-dimensional offset in the current session according to the acupoint coordinate adjustment instruction; taking offset parameters of a global offset correction model as priori, taking the three-dimensional offset as a local observation value, obtaining personalized offset model parameters adaptive to the current object through Bayesian posterior updating fusion, and automatically correcting the three-dimensional acupoint coordinates based on the personalized offset model parameters. The application combines an artificial intelligence model and manual calibration, compensates errors by using global rules of a universal model, differences in anatomical characteristics of different individuals, and positioning deviation between a virtual world and a real world, and significantly improves the adaptability of acupoint recognition and the accuracy of acupoint positioning.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and in particular to a method, device, equipment and medium for human-machine intelligent collaboration in acupoint positioning. Background Technology

[0002] Current methods for locating acupoints mainly rely on manual measurement and experience-based judgment by the operator, or the use of traditional bone measurement methods combined with anatomical atlases. However, this method is highly dependent on human experience, which not only makes it subjective and prone to large positioning errors, but also results in low operational efficiency.

[0003] Therefore, how to achieve high-precision, personalized, and interpretable acupoint positioning in real three-dimensional space is a technical problem that urgently needs to be solved. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to provide a method, device, equipment and medium for human-machine intelligent collaboration in acupoint positioning, which solves the problems of insufficient accuracy, lack of personalization and interpretability in acupoint positioning in the prior art.

[0005] To address the aforementioned technical problems, this invention provides a human-machine intelligent collaborative method for acupoint positioning, comprising: The three-dimensional acupoint coordinates of the current object are obtained through an artificial intelligence positioning model and output through an interactive visualization interface; When the interactive visualization interface detects that an acupoint coordinate adjustment command has been generated, the corresponding three-dimensional offset for dynamic adjustment in this session is obtained according to the acupoint coordinate adjustment command. Using the offset parameters of the global offset correction model as a priori and the three-dimensional offset of the current session as the local observation, personalized offset model parameters adapted to the current object are obtained through Bayesian posterior update fusion, and the three-dimensional acupoint coordinates are automatically corrected based on the personalized offset model parameters.

[0006] Optionally, when the interactive visualization interface detects that an acupoint coordinate adjustment command has been generated, after obtaining the corresponding three-dimensional offset for dynamic adjustment in this session based on the acupoint coordinate adjustment command, the method further includes: The three-dimensional offset is stored in the session cache corresponding to the current session using a key-value pair structure. The session cache is a temporary memory storage medium and supports selective clearing or persistence after the session is closed. The key in the key-value pair is a unique identifier for the acupoint. The value in the key-value pair is a cumulative three-dimensional offset pattern. The cumulative three-dimensional offset pattern includes at least the three-dimensional offset of a single adjustment and the total three-dimensional offset automatically accumulated after multiple adjustments to the same acupoint. Accordingly, using the offset parameters of the global offset correction model as a priori and the three-dimensional offset of the current session as the local observation, personalized offset model parameters adapted to the current object are obtained through Bayesian posterior update fusion. Based on these personalized offset model parameters, the three-dimensional acupoint coordinates are automatically corrected, including: Calculate statistical characteristic values ​​based on the cumulative three-dimensional offset pattern, including at least one of the average value and proportional relationship; Using the offset parameters of the global offset correction model as priors and the statistical feature values ​​as local observations, personalized offset model parameters adapted to the current object are obtained through Bayesian posterior update fusion, and the coordinates of the three-dimensional acupoints are automatically corrected based on the personalized offset model parameters.

[0007] Optionally, when the interactive visualization interface detects that an acupoint coordinate adjustment command has been generated, after obtaining the corresponding three-dimensional offset for dynamic adjustment in this session based on the acupoint coordinate adjustment command, the method further includes: The personalized offset model parameters are anonymized and differential privacy noise is added to obtain anonymized parameters, which are then uploaded to the server. The server then uses the FedAvg algorithm to aggregate and update the global offset correction model based on the anonymized parameters uploaded by each client, and distributes the updated global offset correction model parameters to each client.

[0008] Optionally, after using the offset parameters of the global offset correction model as a priori, the three-dimensional offset of the current session as the local observation, and obtaining personalized offset model parameters adapted to the current object through Bayesian posterior update fusion, and automatically correcting the three-dimensional acupoint coordinates based on the personalized offset model parameters, the method further includes: By statistically analyzing the offset of multiple adjusted acupoints within the same anatomical region, common offset patterns of the anatomical region are obtained, and these common offset patterns are automatically applied to other unadjusted related acupoints within the same anatomical region.

[0009] Optionally, after obtaining the three-dimensional acupoint coordinates of the current object through an artificial intelligence positioning model and outputting them through an interactive visualization interface, the method further includes: The artificial intelligence positioning model outputs nearby candidate points around each acupoint. The clustering module is used to cluster the neighboring candidate points to obtain clusters; Accordingly, the three-dimensional acupoint coordinates of the current object are obtained through an artificial intelligence positioning model and output through an interactive visualization interface, including: The three-dimensional acupoint coordinates and the coordinates of the center point corresponding to the cluster are output through an interactive visualization interface.

[0010] Optionally, the three-dimensional acupoint coordinates of the current object are obtained through an artificial intelligence positioning model and output through an interactive visualization interface, including: The 3D point cloud of the current object is acquired using an RGB-D camera, and based on the 3D point cloud, 3D feature points are acquired using a 3D feature point detection model. Interpolation points are obtained by performing interpolation processing based on the feature points of the three-dimensional human body parts. Based on the interpolation points and the three-dimensional feature points, the coordinates of the three-dimensional acupoints are determined using the adaptive multi-segment bone measurement method and meridian database mapping; the meridian database includes a standardized acupoint location table.

[0011] Optionally, the three-dimensional acupoint coordinates of the current object are obtained through an artificial intelligence positioning model and output through an interactive visualization interface, including: Based on the three-dimensional acupoint coordinates, the locations are marked on the color map of the interactive visualization interface, and the acupoint location reliability and suggested acupoint depth output by the artificial intelligence positioning model are overlaid and displayed.

[0012] The present invention also provides an acupoint positioning human-machine intelligent collaborative device, comprising: The interface output module is used to obtain the three-dimensional acupoint coordinates of the current object through an artificial intelligence positioning model and output them through an interactive visual interface. The interaction module is used to obtain the corresponding three-dimensional offset for dynamic adjustment in this session based on the acupoint coordinate adjustment command when the interactive visualization interface generates an acupoint coordinate adjustment command. The correction module is used to obtain personalized offset model parameters adapted to the current object by using the offset parameters of the global offset correction model as a priori and the three-dimensional offset of the current session as the local observation value, through Bayesian posterior update fusion, and automatically correct the three-dimensional acupoint coordinates based on the personalized offset model parameters.

[0013] This invention also provides an acupoint positioning human-machine intelligent collaborative device, comprising: Memory, used to store computer programs; A processor is used to implement the acupoint positioning human-machine intelligent collaborative method as described above when executing the computer program.

[0014] The present invention also provides a medium storing computer-executable instructions, which, when loaded and executed by a processor, realize the acupoint positioning human-machine intelligent collaborative method as described above.

[0015] The present invention also provides a computer program product, including a computer program / instructions, which, when executed by a processor, implement the steps of the above-described acupoint positioning human-machine intelligent collaborative method.

[0016] As can be seen from the above technical solution, this invention obtains the three-dimensional acupoint coordinates of the current object through an artificial intelligence positioning model and outputs them through an interactive visualization interface. When an acupoint coordinate adjustment command is detected by the interactive visualization interface, the corresponding three-dimensional offset is dynamically adjusted according to the acupoint coordinate adjustment command. Using the offset parameters of the global offset correction model as a priori and the three-dimensional offset of the current session as the local observation value, personalized offset model parameters adapted to the current object are obtained through Bayesian posterior update fusion, and the three-dimensional acupoint coordinates are automatically corrected based on the personalized offset model parameters. The beneficial effects of this invention are: this invention automatically obtains the three-dimensional acupoint coordinates through an artificial intelligence model and relies on an interactive visualization interface to achieve intuitive presentation and manual adjustment, constructing an efficient human-machine collaborative positioning mode; and using the global offset model as a priori and the current adjustment data as the local observation value, personalized offset model parameters are generated through Bayesian posterior update fusion and the acupoint coordinates are automatically corrected. It leverages the advantages of AI (artificial intelligence) models for rapid localization while incorporating the experience-based judgment of precise human calibration, achieving an organic combination of machine intelligence and human intervention. Furthermore, compared to traditional fixed models or single experience-based localization methods, it can fully utilize the global rules of general models while taking into account individual anatomical differences and actual localization deviations, significantly improving the adaptability and localization accuracy of acupoint coordinates, reducing the cost of repeated manual adjustments, and achieving intelligent, personalized, and high-precision acupoint localization.

[0017] In addition, the present invention also provides an acupoint positioning human-machine intelligent collaborative device, equipment and medium, which also have the above-mentioned beneficial effects. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0019] Figure 1 A flowchart of a human-machine intelligent collaborative method for acupoint positioning provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of an acupoint positioning human-machine intelligent collaborative device provided in an embodiment of the present invention; Figure 3This is a schematic diagram of the structure of an acupoint positioning human-machine intelligent collaborative device provided in an embodiment of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Please refer to Figure 1 , Figure 1 A flowchart illustrating a human-machine intelligent collaborative method for acupoint location according to an embodiment of the present invention. The method may include: S101: Obtain the three-dimensional acupoint coordinates of the current object through an artificial intelligence positioning model and output them through an interactive visualization interface.

[0022] The steps in this embodiment can be executed by a designated electronic device, which can be a server, a portable terminal, or other forms. In this embodiment, the executing entity is an acupoint positioning human-machine intelligent collaborative device. This embodiment does not specifically limit the artificial intelligence positioning model, as long as it can achieve three-dimensional acupoint coordinate positioning. The interactive visualization interface in this embodiment refers to a human-computer interaction visualization interface.

[0023] Furthermore, to improve the accuracy of three-dimensional acupoint coordinate positioning, the interface for obtaining the three-dimensional acupoint coordinates of the current object through the aforementioned artificial intelligence positioning model may include: Step 11: Use an RGB-D camera to acquire the 3D point cloud of the current object.

[0024] An RGB-D camera is a device that can simultaneously capture color images (RGB) and depth information (D stands for Depth). Before using an RGB-D camera, calibration is required. Camera calibration involves photographing a calibration board of known size (e.g., a checkerboard) to calculate the precise alignment between the color and depth images. Camera calibration prevents misalignment of color and distance data. After calibration, the depth map is converted into a large number of 3D points, forming a "point cloud." Each point cloud has precise (x, y, z) coordinates (x for left / right, y for up / down, and z for front / back / depth). The 3D point cloud formation process involves converting the 2D depth map into a 3D point cloud using the camera's intrinsic and extrinsic parameter matrices.

[0025] Step 12: Based on the 3D point cloud, obtain the 3D feature points using the 3D feature point detection model.

[0026] In this step, the 3D feature point detection model can employ the HRNet-3D model. This model is a high-resolution neural network specifically designed to detect key points such as joints and skeletal landmarks. It automatically identifies 3D feature points on a point cloud or color image. Examples include key points for various body parts (such as the nose, eyes, ears, shoulders, elbows, wrists, hips, knees, and ankles, including acromion, anterior superior iliac spine, and spinal key points) and traditional Chinese medicine feature points (such as spinal key points, iliac crest / anterior superior iliac spine, etc.). For instance, during 3D feature point detection, 2D feature points can be obtained first through color image detection, and then depth information can be fused to obtain 3D feature points.

[0027] Step 13: Perform interpolation processing based on the feature points of the three-dimensional human body parts to obtain interpolation points.

[0028] After detecting the feature points of three-dimensional human body parts, interpolation is needed between these points to estimate the location of acupoints. This embodiment specifically introduces cubic spline interpolation: using piecewise cubic polynomial curves to ensure continuity in position, slope, and curvature. The cubic spline interpolation method uses a polynomial in each interval to ensure that the values ​​at the endpoints, the first derivative, and the second derivative are equal for adjacent segments. Ultimately, the precise coordinates of the acupoint on the curve can be output.

[0029] Step 14: Based on the interpolation points and three-dimensional feature points, the coordinates of three-dimensional acupoints are determined using the adaptive multi-segment bone measurement method and meridian database mapping; the meridian database includes a standardized acupoint location table.

[0030] Traditional Chinese medicine's "bone measurement method" uses the width of the user's own fingers or the length of their bones as the unit of measurement. However, this method does not take into account individual differences. Therefore, this step adopts an adaptive multi-segment bone measurement method: first, the length of multiple bone segments throughout the body (one segment in the head, multiple segments in the upper limbs, multiple segments in the lower limbs, etc.) is measured, and the "measurement" ratio of each segment is dynamically calculated (adapting to different body types and heights). Then, it is mapped to a meridian database to finally generate three-dimensional acupoint coordinates. Precise location points in three-dimensional space. The standardized acupoint location table pre-stored in this meridian database includes the fourteen meridians and the relative bone position of each acupoint.

[0031] Furthermore, the AI-powered localization model can also output confidence scores, and / or, when the acupoint is used in acupuncture / massage, a recommended acupoint depth. This allows us to obtain a tuple for each acupoint output by the AI-powered localization model: three-dimensional acupoint coordinates. Confidence score and recommended acupoint depth.

[0032] Specifically, AI-powered positioning models sometimes misjudge situations due to low lighting or complex skin texture. Therefore, a confidence score is introduced to help users determine the reliability of the results and avoid blind operations. The confidence score is a score between 0 and 1 (1 being completely reliable), typically derived from the output probability of the AI ​​positioning model or from multi-source data fusion (accuracy of 3D human body feature point detection, spline interpolation error, and meridian database matching degree).

[0033] Because acupoints are a three-dimensional concept, not a two-dimensional one, the "depth" of an acupoint refers to its distance from the skin. Specifically, when pressing an acupoint with a finger or tool, "depth" refers to the degree to which the soft tissue sinks under pressure, producing a feeling of soreness and distension without causing severe pain. Recommended acupoint depths can be calculated by combining standard depths from a database (e.g., 0.5-1 cun for Hegu acupoint), the user's body thickness (inferred from the z-value), and a safety margin to arrive at a suggested value (unit: centimeters or inches). Different individuals' BMI (weight), gender (different body fat distribution between men and women), and age (young people have tighter skin, older people have looser skin and muscle atrophy) can lead to significant differences in the actual acupoint depth required for the same acupoint. Using a uniform value might result in a depth that is too shallow and ineffective, or too deep and damaging to the tissue. Therefore, the recommended acupoint depth output by the AI ​​positioning model can be referenced using the following formula: Recommended acupoint depth = base_depth × scale(BMI, gender, age). An upper limit constraint is used to ensure safety. For example, the upper limit constraint could be: the final result should not exceed 1.2% (i.e., 1.2 times) of the base depth. Here, `base_depth` is a default value derived from standard anatomical data, clinical guidelines, or average models, without considering individual differences. `scale(BMI, gender, age)` is a scaling factor, a number greater than 0 (usually between 0.8 and 1.3), dynamically calculated based on the specific parameters of the current object. BMI (Body Mass Index): BMI = weight (kg) ÷ [height (m)]². Overweight individuals (high BMI) have thicker subcutaneous fat, requiring a deeper scale; thinner individuals (lower BMI) may require a shallower scale. Therefore, `scale` increases appropriately with BMI. Gender: Men typically have a higher muscle mass and less fat, so `scale` may be close to 1.0 or slightly higher; women have a relatively thicker fat layer, so `scale` may be slightly lower or adjusted according to acupoints. Age: Young people (18-35 years old) have good skin elasticity and firm muscles, so `scale` ≈ 1.0; middle-aged and elderly people have looser skin and changes in fat distribution, so `scale` needs fine-tuning (usually slightly smaller to prevent over-deepening).

[0034] For example, the specific calculation of scale is: scale = 1.0 + α × (BMI - 22) + β × gender_factor + γ × age_factor. Where α, β, and γ are coefficients for fitting clinical data; for example, α = 0.01 means that for every 1 increase in BMI, the depth increases by 1%; gender_factor = 0 for males, -0.05 for females, etc. Specific coefficients can be determined through clinical trials. A 120% safety constraint is applied: for example, if base_depth = 20mm, and the calculated temp_depth = 26mm (130%), then a forced value of 24mm (120%) is applied. The final output recommended depth is a safe, personalized value (e.g., "Recommended depth: 22.4 mm").

[0035] To make the viewing experience more intuitive for users, the acupoint tuples output by the above model are displayed on an interactive visualization interface with acupoint locations and recommended depth overlaid on a color image. The confidence score is indicated by color: using the color image as a background, "confidence spheres" are overlaid at the corresponding positions: green sphere: confidence > 0.9 (very reliable); yellow sphere: 0.7 ≤ confidence ≤ 0.9 (moderately reliable, caution advised); red sphere: confidence < 0.7 (low reliability, manual adjustment may be required). The size of the spheres can vary with the confidence level or depth, and the transparency indicates the level of detail.

[0036] Furthermore, after obtaining the three-dimensional acupoint coordinates of the current object through the artificial intelligence positioning model and outputting them through an interactive visualization interface, it may also include: outputting the neighboring candidate points around each acupoint through the artificial intelligence positioning model; clustering the neighboring candidate points to obtain clusters; correspondingly, obtaining the three-dimensional acupoint coordinates of the current object through the artificial intelligence positioning model may include: outputting the three-dimensional acupoint coordinates and the coordinates of the center point corresponding to the cluster through an interactive visualization interface.

[0037] In this embodiment, the artificial intelligence positioning model will also output k candidate point clusters. After the acupoint coordinates are determined, the neighboring candidate points around the acupoint are grouped according to spatial distance using a clustering algorithm (such as K-means or DBSCAN). The k most representative clusters can be selected (e.g., k=3). Each cluster may be the center point or a small set of points, which are output as candidates to the interactive visualization interface to deal with detection noise or body shape differences.

[0038] S102: When an acupoint coordinate adjustment command is detected by the interactive visualization interface, the corresponding three-dimensional offset for dynamic adjustment in this session is obtained according to the acupoint coordinate adjustment command.

[0039] It should be noted that the acupoint coordinate adjustment command in this embodiment refers to the adjustment command issued by the operator through the interactive visual interface. The adjustment command can be a touch command or a voice command. The specific content of the adjustment command can be the acupoint coordinate position, depth, etc. In addition, the interactive visual interface supports adjustment commands generated by drag, voice command and area selection. For example, (1) the user (such as an experienced operator) directly selects the ball, drags to adjust the position, selects the candidate cluster and modifies the depth on the interactive visual interface. Taking drag adjustment of position as an example: drag the marker point representing the acupoint directly on the interactive visual interface with your finger, and immediately update the 3D coordinates of the acupoint by capturing the finger movement trajectory in real time (the update frequency is usually tens of times per second). Taking confirmation or area selection adjustment as an example: you can click the "confirm" button to indicate that the current acupoint position is correct, or draw a circle / box on the screen with your finger to select an area, then all acupoints in the selected area will be regarded as the same group, and subsequent adjustments or marking as verified will be applied uniformly. (2) The user responds in real time through voice recognition by natural language commands on the interactive visual interface. For example, commands like "adjust upwards by 2mm," "move to the right by 1.5mm," and "advance forward by 0.5mm" can be accurately converted from speech to text using acupoint positioning human-machine intelligent collaborative device. The device then analyzes the commands: "upwards" corresponds to a specific axis in the 3D coordinate system (usually the positive Z-axis is the upward / depth direction); "2mm" is converted to a specific numerical value; other directions (left / right, front / back) correspond to the X-axis or Y-axis. The analysis result automatically generates a three-dimensional offset delta.

[0040] Upon receiving an acupoint coordinate adjustment command, the system automatically records the three-dimensional offset (i.e., the change), rather than the absolute position, for easier subsequent accumulation and analysis. The recording format is as follows: .in: This is the offset in the X-axis direction (unit: mm; a positive value usually indicates to the right). Y-axis offset (unit: mm, positive values ​​usually indicate forward / towards the user's head); This represents the offset along the Z-axis (unit: mm; a positive value usually indicates upward / increased depth). If the three-dimensional acupoint coordinates... , These are the initial X, Y, and Z positions of the acupoint in the current object's body coordinate system, in mm.

[0041] After adjustment, the optimized coordinate calculation formula is as follows: .

[0042] To ensure rapid response and avoid lag for users, multithreading can be used to divide rendering, coordinate calculation, and UI updates into multiple parallel tasks, preventing the main thread from blocking. Specifically, GPU (Graphics Processing Unit) rendering can be used to recalculate the optimized coordinates of all acupoints that need adjustment. The system then redraws and overlays images on the screen in real time (e.g., red markers, labels, depth indicators, etc.). Alternatively, to make the results more intuitive and user-friendly, it can support real-time augmented reality (AR) overlays: using the Unity framework (cross-platform compatible with Android / iOS / Windows) or ARKit (for Apple devices), it overlays on the live camera feed: 3D coordinate points (virtual spheres or crosshairs, whose positions update in real time with (x,y,z)), confidence scores (e.g., a green "92%)", recommended acupoint depth values, and virtual needle guide lines (extending from the skin surface to the recommended depth). As the camera moves, the markers follow the skin without noticeable delay. Compared to the unintuitive nature of simply displaying numbers, introducing AR technology significantly improves operational accuracy and safety, allowing users to see the virtual needle's position on their real body.

[0043] Furthermore, when an acupoint coordinate adjustment command is detected by the interactive visualization interface, after obtaining the corresponding three-dimensional offset for dynamic adjustment in this session based on the acupoint coordinate adjustment command, the method may further include: storing the three-dimensional offset in the session cache corresponding to this session using a key-value pair structure. The session cache is a temporary memory storage medium and supports selective clearing or persistence after the session is closed. The key in the key-value pair is a unique identifier for the acupoint; the value in the key-value pair is the cumulative three-dimensional offset pattern; the cumulative three-dimensional offset pattern includes at least the three-dimensional offset of a single adjustment and the total three-dimensional offset automatically accumulated after multiple adjustments of the same acupoint; correspondingly, a global offset correction is used. Using the offset parameters of the positive model as priors and the 3D offset of the current session as local observations, personalized offset model parameters adapted to the current object are obtained through Bayesian posterior update fusion. The 3D acupoint coordinates are then automatically corrected based on the personalized offset model parameters. This can include: calculating statistical feature values ​​containing at least one of average value and proportional relationship based on the cumulative 3D offset pattern; using the offset parameters of the global offset correction model as priors and the statistical feature values ​​as local observations, personalized offset model parameters adapted to the current object are obtained through Bayesian posterior update fusion, and the 3D acupoint coordinates are then automatically corrected based on the personalized offset model parameters.

[0044] In this embodiment, the three-dimensional offset delta is stored in the session cache. The session cache refers to the temporary memory storage of the current session. After the session is closed, it can be cleared or persisted. Its storage format adopts key-value pair (key-value structure). Among them, the key is the unique ID of the acupoint (e.g., "ST36" Zusanli acupoint, the user number and acupoint name, or the UUID generated by the system); the value is the cumulative three-dimensional offset pattern (record the delta this time, and adjust the cumulative or extraction mode with the calibration experience of the historical operator). Among them, the meaning of the cumulative three-dimensional offset pattern is: (1) Single adjustment: direct storage (2) If the same acupoint is adjusted multiple times: automatic accumulation can be performed. For example, the first delta1=(1,0,0), the second delta2=(0,2,0), and the total three-dimensional offset after accumulation =(1,2,0); (3) Extract the three-dimensional offset pattern: calculate the average value, proportional relationship, etc., to prepare for subsequent automatic inference.

[0045] S103: Using the offset parameters of the global offset correction model as a priori and the 3D offset of the current session as the local observation, personalized offset model parameters adapted to the current object are obtained through Bayesian posterior update fusion, and the 3D acupoint coordinates are automatically corrected based on the personalized offset model parameters.

[0046] This step first uses existing global offset correction model parameters as general prior knowledge, then uses the 3D offset of acupoints obtained in this human-computer interaction as actual observation data. A Bayesian posterior update algorithm is used to fuse the two, generating personalized offset parameters adapted to the current object's body characteristics. Finally, based on these parameters, the 3D acupoint coordinates identified by the AI ​​positioning model are automatically and accurately corrected. This approach preserves the stability of the general model while also considering individual differences and the practical needs of manual calibration. After automatically correcting the acupoint coordinates, this embodiment allows for subsequent acupuncture, massage, or fumigation treatments. Alternatively, the aforementioned acupoint identification and positioning results can be further applied to beauty and body care devices, such as smart massagers or LED (phototherapy) masks. These devices can automatically and accurately locate acupoints on the face or body and plan personalized anti-aging massage paths based on the acupoints. They can also be integrated into fitness motion capture systems, introducing acupoints as important anatomical landmarks outside of joints into the analysis, evaluation, and measurement systems. This significantly improves the overall effect of athlete posture analysis and the accuracy of biomechanical measurements, thereby providing acupoint-based exercise recovery guidance. Furthermore, the acupoint positioning human-machine intelligent collaborative method provided in this embodiment can also be embedded in smart home massage chairs, service robots, and various wearable devices to achieve automatic acupoint scanning through clothing and non-invasive massage or electrical stimulation path planning. In addition, by overlaying meridian and acupoint information onto a virtual three-dimensional human body model through an AR (Augmented Reality) / VR (Virtual Reality) interactive platform, it can be widely used for public health education, entertainment games, and immersive relaxation experiences. Therefore, users can enjoy personalized, non-invasive acupoint services and experiences in fields such as beauty and wellness, motion capture, sports recovery, smart home, health education, and digital entertainment without having to visit professional institutions.

[0047] Furthermore, when an acupoint coordinate adjustment command is detected by the interactive visualization interface, after obtaining the corresponding three-dimensional offset for dynamic adjustment in this session based on the acupoint coordinate adjustment command, the process may further include: anonymizing the personalized offset model parameters and adding differential privacy noise to obtain anonymized parameters, which are then uploaded to the server; so that the server updates the global offset correction model based on the anonymized parameters uploaded by each client using the FedAvg algorithm, and distributes the updated global offset correction model parameters to each client.

[0048] To further enhance the model's generalization ability and privacy security, after updating the local personalized offset parameters based on the 3D offset of this session, this embodiment further introduces a privacy protection and federated collaborative optimization mechanism. Personalized data generated in a single session has significant optimization value for acupoint positioning accuracy, but directly uploading raw parameters (such as vital signs, symptoms, and lifestyle habits provided by the current object in a single session) can easily lead to individual privacy leaks. Therefore, a Bayesian posterior update is first performed locally to quickly achieve personalized adaptation within the session. Then, the updated personalized offset model parameters are anonymized, and differential privacy noise is added to eliminate privacy-related information while preserving parameter validity to the maximum extent. The anonymized parameters are then uploaded to the server. After receiving the anonymized parameters uploaded by each client, the server uses the FedAvg algorithm for secure aggregation, weighting the local model parameters (i.e., personalized offset model parameters) of multiple institutions with data volume as the weight, iteratively updating the global offset correction model, and distributing the optimized global model parameters to each client to achieve multi-center collaborative iteration.

[0049] The entire optimization process is divided into three logically coherent stages, taking into account the real-time positioning, individual accuracy and data privacy and security: (1) The first stage is local real-time update within the session. Using the global offset correction model parameters as a priori, the Bayesian posterior update is performed in combination with the three-dimensional offset generated in this session. Under the premise of no retraining, personalized offset parameters adapted to the current object are quickly generated, which not only retains the general rules of the population, but also accurately integrates individual characteristics. All calculations are completed locally, and the original data does not leave the local environment. For example, let the average offset parameter of the global offset correction model be δ_global, and a local temporary delta is calculated for the new data of this session, namely the personalized offset parameter δ_local. Then the delta after the Bayesian posterior update is: δ_posterior=(1-α)×δ_global+α×δ_local. Wherein, α is the learning rate (between 0 and 1, controlling the influence weight of this data, usually dynamically adjusted according to the amount of data). (2) The second stage is privacy protection processing and secure upload. To avoid inferring individual information from parameters, all identifying information is first stripped away (e.g., removing the user ID and all direct identifiers: deleting name, ID number, medical ID, precise timestamp, etc.). Privacy is then enhanced by adding Gaussian or Laplace difference privacy noise, such as adding noise to δ_posterior: δ_anonymized = δ_posterior + ... (0, σ²). Among them, σ is calculated based on the privacy budget ε (the smaller the ε, the stronger the privacy) and data sensitivity; if necessary, quantization compression is used to convert floating-point parameters into low-precision format to further reduce the risk of information leakage. In this way, by uploading only the anonymized parameters after desensitization (such as δ_anonymized, corresponding data volume n_k, local training rounds and other statistical information), the original physical characteristics and privacy data are prevented from being transmitted. (3) The third stage is the offline federated aggregation update based on FedAvg. Each user institution (including multiple clients in each user institution) only uploads the privacy-protected model parameters. The server performs weighted averaging based on the proportion of data volume of each client, completes the iterative optimization of the global offset correction model, and sends the updated global model back to the terminal of each user institution. In response to the problem of significant differences in positioning deviation among different body types, subgroups can be divided according to dimensions such as BMI, and federated fine-tuning can be performed separately on the subgroup data to obtain a dedicated sub-model to further improve the positioning accuracy of the subdivided population. This approach enables multi-institutional collaborative optimization of the model while strictly protecting data privacy and security. It solves the problem of insufficient data samples from a single institution and avoids the leakage of individual information, enabling continuous iterative optimization of the global acupoint positioning model while making local personalized positioning more accurate and reliable.

[0050] Detailed description of the FedAvg mathematical process: 1. Symbol definition: K: The total number of clients participating in this federated learning; n k : Size of the local dataset of the kth client (number of user samples); n = ∑ K {k=1} n k The total amount of data from all clients, from the first to the Kth. θ k {t}: Personalized offset model parameters (weight vector) for the kth client at the start of the tth round of global iteration; θ k {t+1}: Updated personalized offset model parameters after the kth client completes several rounds of training locally; θ global {t}: The current global offset correction model parameters at the start of the t-th round of global iteration; θ global {t+1}: Global offset correction model parameters after the t-th round of global iteration.

[0051] 2. Local training process: Each client uses its own local data and the current global offset correction model θ globalUsing {t} as the initial parameter, perform multiple rounds (usually E rounds) of stochastic gradient descent (SGD) training locally to obtain the locally updated parameter θ. k {t+1}. The local data remains completely unchanged, and the training process is entirely localized.

[0052] 3. Global aggregation update formula (FedAvg core formula): θ global {t+1} = ∑ K {k=1} n k / n·θ k {t+1}. The personalized offset model parameters θ for each client, from the 1st to the Kth. k {t+1}, according to its data size n k The proportion of data points (n) to the total data volume is used as the weight for a weighted sum. Institutions with larger data volumes contribute more weight and have a greater impact on the global offset correction model. Institutions with smaller data volumes have smaller weights to avoid being overly influenced.

[0053] 4. Complete iterative process: Round 0: Server initialization θ global {0} and distribute it to all K clients. For each round t = 0, 1, 2, ...; a. Each client receives θ global {t}; b. Each client uses its own n locally. k Train the sample for several rounds to obtain θ k {t+1}; c. Each client uploads θ k {t+1} (anonymized); d. The server calculates the weighted average: θ global {t+1} = ∑ K {k=1} n k / n·θ k {t+1}; (weighted average from the 1st to the Kth) e. Distribute the new global offset correction model back to the client to start the next round, repeating until convergence or the preset number of rounds is reached.

[0054] 5. Privacy protection advantages: The raw data remains local; only model parameters are transmitted, and this is further protected by differential privacy noise. Even if an attacker obtains all uploaded parameters, it is difficult to reconstruct the data of a single user.

[0055] Targeted Optimization: Fine-tuning of Body Size Sub-models. In practical applications, we found significant differences in physiological characteristics among different body size groups (e.g., obese subgroups with BMI > 28), making it difficult for the global offset correction model to optimize all groups simultaneously. Therefore, we allow for individual FedAvg fine-tuning for specific subgroups: a subset of local data with BMI > 28 is selected; the FedAvg process described above is run on this subset to obtain a dedicated sub-model θ_sub; this sub-model can be combined with the global offset correction model (e.g., weighted fusion or conditional invocation). This further improves personalized accuracy.

[0056] Furthermore, after using the offset parameters of the global offset correction model as a priori, the three-dimensional offset of the current session as the local observation value, and obtaining personalized offset model parameters adapted to the current object through Bayesian posterior update fusion, and automatically correcting the three-dimensional acupoint coordinates based on the personalized offset model parameters, it can also include: statistically analyzing the offsets of multiple adjusted acupoints in the same anatomical region to obtain the common offset patterns of the anatomical region, and automatically applying the common offset patterns to other unadjusted related acupoints in the same anatomical region.

[0057] It should be noted that multiple related acupoints are usually adjusted simultaneously when adjusting acupoints. For example, acupoints on the entire arm. The acupoint coordinate adjustment command identifies the acupoint to be corrected. Based on the body part to which the acupoint belongs (same limb: such as the left forearm, right calf, etc., determined by predefined anatomical grouping), the offset ratio or similarity of the adjustment is calculated (e.g., the average X component of all adjusted deltas is +1.2mm), and automatically applied to the same group of acupoints that have not been adjusted or have initially deviated in positioning. For example, the acupoints to be corrected are three acupoints on the arm: Acupoint A: Acupoint B: Acupoint C: Calculate the common offset patterns: average dx ≈ 1.37mm, average dy ≈ 0.1mm, average dz ≈ 0.5mm or a scaling factor k (relative to the reference delta), and then apply the offset to other acupoints D on the same limb: Final Update .

[0058] Furthermore, mode switching and conflict decision-making can be added. Specifically, (1) collect basic data, including: operator's years of experience, historical adjustment rate, three-dimensional acupoint coordinates (AI_pos) and confidence (range 0~1, 0.9 indicates very reliable) and three-dimensional offset (delta) output by the artificial intelligence positioning model. (2) Determine whether to enter "forced adjustment mode" (operators with insufficient experience or those who frequently make modifications must manually confirm). If the preset conditions are met (e.g., operator experience < 5 years or historical adjustment rate > 20%), then the forced adjustment mode is forcibly entered: the artificial intelligence positioning model outputs three-dimensional acupoint coordinates and confidence, and the operator must view, intervene and finally confirm the final position. If the preset conditions are not met, then determine whether to enter "supervised mode" (skilled operators can pass automatically). For example, when the operator has ≥ 5 years of experience and the historical adjustment rate is ≤ 20%, further checks are performed: if the confidence level is > 0.9 or the historical adjustment rate is < 10%, then switch to the supervision mode: the three-dimensional acupoint coordinates output by the artificial intelligence positioning model are automatically passed and directly used in the subsequent treatment process. The operator can still view (supervision) in real time, but no manual confirmation is required to execute. Record whether the operator manually modifies it later as future adjustment rate statistics. (3) Regardless of the mode, when the operator puts forward his / her own adjustment opinion, the difference needs to be calculated: difference = |operator's expected position - AI_pos|, where the operator's expected position = AI_pos + delta). If the difference is ≤ 0.5 cm, it is considered as "small conflict" or "no conflict", and the operator's opinion or AI suggestion (depending on the mode) is directly adopted without triggering weighted fusion. If the difference is > 0.5 cm, it is considered as "significant conflict", and then it needs to enter the weighted fusion decision: Both 0.7 and 0.3 are weighting coefficients. 0.7 indicates that the 3D acupoint coordinates output by the AI ​​positioning model account for 70% of the final result, while the operator's correction accounts for 30%. The system outputs the final acupoint position (final_pos), displays the current mode (forced adjustment / supervised adjustment), whether weighted fusion was used, and the reason for the weighting; records whether the operator accepted the final result, used to update the "historical adjustment rate"; and generates an auditable log.

[0059] Furthermore, considering that the output of the three-dimensional acupoint coordinates by the artificial intelligence positioning model may be affected by noise (camera shake, lighting changes, skin curvature, etc.), optimization algorithms (such as Kalman filtering, Bundle Adjustment, or deep learning post-processing) can be used to further optimize the coordinates.

[0060] The acupoint positioning human-machine intelligent collaborative method provided in this embodiment of the invention proceeds as follows: S101: The three-dimensional acupoint coordinates of the current object are obtained through an artificial intelligence positioning model and output through an interactive visualization interface; S102: When an acupoint coordinate adjustment command is detected by the interactive visualization interface, the corresponding three-dimensional offset for dynamic adjustment in this session is obtained according to the acupoint coordinate adjustment command; S103: Using the offset parameters of the global offset correction model as a priori and the three-dimensional offset of the current session as the local observation value, personalized offset model parameters adapted to the current object are obtained through Bayesian posterior update fusion, and the three-dimensional acupoint coordinates are automatically corrected based on the personalized offset model parameters. This method automatically obtains the three-dimensional acupoint coordinates through an artificial intelligence model and relies on an interactive visualization interface to achieve intuitive presentation and manual adjustment, thus constructing an efficient human-machine collaborative positioning mode; and using the global offset model as a priori and the current adjustment data as the local observation value, personalized offset model parameters are generated through Bayesian posterior update fusion and the acupoint coordinates are automatically corrected. It leverages the advantages of AI models for rapid localization while incorporating the experience-based judgment of precise human calibration, achieving an organic combination of machine intelligence and human intervention. Furthermore, compared to traditional fixed models or single experience-based localization methods, it can fully utilize the global rules of general models while taking into account individual anatomical differences and actual localization deviations, significantly improving the adaptability and positioning accuracy of acupoint coordinates, reducing the cost of repeated manual adjustments, and achieving intelligent, personalized, and high-precision acupoint localization.

[0061] The following describes the acupoint positioning human-machine intelligent collaborative device provided in the embodiments of the present invention. The acupoint positioning human-machine intelligent collaborative device described below can be referred to in correspondence with the acupoint positioning human-machine intelligent collaborative method described above.

[0062] Please refer to the details. Figure 2 , Figure 2 A schematic diagram of a positioning and human-machine intelligent collaborative device provided in an embodiment of the present invention may include: The interface output module 100 is used to obtain the three-dimensional acupoint coordinates of the current object through an artificial intelligence positioning model and output them through an interactive visual interface. The interaction module 200 is used to obtain the corresponding three-dimensional offset for dynamic adjustment in this session according to the acupoint coordinate adjustment command when the interactive visualization interface generates an acupoint coordinate adjustment command. The correction module 300 is used to obtain personalized offset model parameters adapted to the current object by using the offset parameters of the global offset correction model as a priori and the three-dimensional offset of the current session as the local observation value through Bayesian posterior update fusion, and automatically correct the three-dimensional acupoint coordinates based on the personalized offset model parameters.

[0063] Furthermore, based on the above embodiments, the acupoint positioning human-machine intelligent collaborative device may further include: The caching module is used to, when the interactive visualization interface detects an acupoint coordinate adjustment command, obtain the corresponding three-dimensional offset for dynamic adjustment in the current session based on the acupoint coordinate adjustment command, and then store the three-dimensional offset in the session cache corresponding to the current session using a key-value pair structure. The session cache is a temporary memory storage medium and supports selective clearing or persistence after the session is closed. The key in the key-value pair is a unique identifier for the acupoint; the value in the key-value pair is a cumulative three-dimensional offset pattern; the cumulative three-dimensional offset pattern includes at least the three-dimensional offset of a single adjustment and the total three-dimensional offset automatically accumulated after multiple adjustments to the same acupoint. Accordingly, the correction module 300 may include: The calculation unit is used to calculate a statistical feature value, including at least one of the average value and the proportional relationship, based on the cumulative three-dimensional offset pattern. The correction unit is used to obtain personalized offset model parameters adapted to the current object by using the offset parameters of the global offset correction model as a priori and the statistical feature values ​​as local observations, through Bayesian posterior update fusion, and to automatically correct the coordinates of the three-dimensional acupoints based on the personalized offset model parameters.

[0064] Furthermore, based on the above embodiments, the acupoint positioning human-machine intelligent collaborative device may further include: The upload module is used to, when the interactive visualization interface generates an acupoint coordinate adjustment command, obtain the corresponding three-dimensional offset for dynamic adjustment in the current session based on the acupoint coordinate adjustment command, anonymize the personalized offset model parameters and add differential privacy noise to obtain anonymized parameters, and upload them to the server; so that the server can aggregate and update the global offset correction model based on the anonymized parameters uploaded by each client using the FedAvg algorithm, and distribute the updated global offset correction model parameters to each client.

[0065] Furthermore, based on the above embodiments, the acupoint positioning human-machine intelligent collaborative device may further include: The correlation correction module is used to obtain personalized offset model parameters adapted to the current object by using the offset parameters of the global offset correction model as a priori and the three-dimensional offset of the current session as the local observation value through Bayesian posterior update fusion. After automatically correcting the three-dimensional acupoint coordinates based on the personalized offset model parameters, the module performs statistical analysis on the offsets of multiple adjusted acupoints in the same anatomical region to obtain the common offset pattern of the anatomical region and automatically applies the common offset pattern to other unadjusted related acupoints in the same anatomical region.

[0066] Furthermore, based on the above embodiments, the acupoint positioning human-machine intelligent collaborative device may further include: The neighboring candidate point acquisition module is used to obtain the three-dimensional acupoint coordinates of the current object through the artificial intelligence positioning model and output them through an interactive visualization interface, and then output the neighboring candidate points around each acupoint through the artificial intelligence positioning model. The clustering module is used to cluster the neighboring candidate points to obtain clusters; Correspondingly, the interface output module 100 may include: The first output unit is used to output the coordinates of the three-dimensional acupoints and the coordinates of the center points corresponding to the clusters through an interactive visualization interface.

[0067] Furthermore, based on the above embodiments, the interface output module 100 may include: The three-dimensional part feature point acquisition unit is used to acquire the three-dimensional point cloud of the current object using an RGB-D camera, and to acquire three-dimensional part feature points based on the three-dimensional point cloud using a three-dimensional part feature point detection model. An interpolation unit is used to perform interpolation processing based on the feature points of the three-dimensional human body parts to obtain interpolation points; The three-dimensional acupoint coordinate determination unit is used to determine the three-dimensional acupoint coordinates based on the interpolation points and the three-dimensional feature points, using an adaptive multi-segment bone measurement method and meridian database mapping; the meridian database includes a standardized acupoint location table.

[0068] Furthermore, based on the above embodiments, the interface output module 100 may include: The second output unit is used to mark the location on the color map of the interactive visualization interface according to the three-dimensional acupoint coordinates, and to overlay and display the acupoint location reliability and suggested acupoint depth output by the artificial intelligence positioning model.

[0069] It should be noted that the order of the modules and units in the above-mentioned acupoint positioning human-machine intelligent collaborative device can be changed without affecting the logic.

[0070] After automatically correcting the acupoint coordinates, this embodiment allows for subsequent acupuncture, massage, or fumigation of the acupoints. Alternatively, the acupoint identification and positioning results can be further applied to beauty and body care devices, such as smart massagers or LED (phototherapy) masks. These devices can automatically and accurately locate acupoints on the face or body and plan personalized anti-aging massage paths based on the acupoints. They can also be integrated into fitness motion capture systems, introducing acupoints as important anatomical landmarks outside of joints into the analysis, evaluation, and measurement systems, significantly improving the overall effect of athlete posture analysis and the accuracy of biomechanical measurements, thereby providing acupoint-based exercise recovery guidance. Furthermore, the acupoint positioning human-machine intelligent collaborative method provided in this embodiment can be embedded in smart home massage chairs, service robots, and various wearable devices to achieve automatic acupoint scanning through clothing and non-invasive massage or electrical stimulation path planning. In addition, by overlaying meridian and acupoint information onto a virtual three-dimensional human body model through an AR (Augmented Reality) / VR (Virtual Reality) interactive platform, it can be widely used for public health education, entertainment games, and immersive relaxation experiences. Therefore, users can enjoy personalized, non-invasive acupoint services and experiences in fields such as beauty and wellness, motion capture, sports recovery, smart home, health education, and digital entertainment without having to visit professional institutions.

[0071] The human-machine intelligent collaborative acupoint positioning device provided in this embodiment of the invention includes an interface output module 100, which acquires the three-dimensional acupoint coordinates of the current object through an artificial intelligence positioning model and outputs them through an interactive visual interface; an interaction module 200, which, when an acupoint coordinate adjustment command is detected by the interactive visual interface, obtains the corresponding three-dimensional offset for dynamic adjustment in the current session according to the acupoint coordinate adjustment command; and a correction module 300, which uses the offset parameters of the global offset correction model as a priori, uses the three-dimensional offset of the current session as the local observation value, and obtains personalized offset model parameters adapted to the current object through Bayesian posterior update fusion, and automatically corrects the three-dimensional acupoint coordinates based on the personalized offset model parameters. This device automatically acquires three-dimensional acupoint coordinates through an artificial intelligence model and relies on an interactive visual interface to achieve intuitive presentation and manual adjustment, constructing an efficient human-machine collaborative positioning mode; and uses the global offset model as a priori and the current adjustment data as the local observation value, generates personalized offset model parameters through Bayesian posterior update fusion and automatically corrects the acupoint coordinates. It leverages the advantages of AI models for rapid localization while incorporating the experience-based judgment of precise human calibration, achieving an organic combination of machine intelligence and human intervention. Furthermore, compared to traditional fixed models or single experience-based localization methods, it can fully utilize the global rules of general models while taking into account individual anatomical differences and actual localization deviations, significantly improving the adaptability and positioning accuracy of acupoint coordinates, reducing the cost of repeated manual adjustments, and achieving intelligent, personalized, and high-precision acupoint localization.

[0072] The following is a description of the acupoint positioning human-machine intelligent collaborative device provided in the embodiments of the present invention. The acupoint positioning human-machine intelligent collaborative device described below can be referred to in correspondence with the acupoint positioning human-machine intelligent collaborative method described above.

[0073] Please refer to Figure 3 , Figure 3 A schematic diagram of the structure of an acupoint positioning human-machine intelligent collaborative device provided in an embodiment of the present invention may include: Memory 10 is used to store computer programs; The processor 20 is used to execute computer programs to implement the above-mentioned acupoint positioning human-machine intelligent collaborative method.

[0074] The memory 10, processor 20, and communication interface 31 all communicate with each other through the communication bus 32.

[0075] In this embodiment of the invention, the memory 10 is used to store one or more programs. The programs may include program code, which includes computer operation instructions. In this embodiment of the invention, the memory 10 may store programs for implementing the following functions: The three-dimensional acupoint coordinates of the current object are obtained through an artificial intelligence positioning model and output through an interactive visualization interface; When the interactive visualization interface detects that an acupoint coordinate adjustment command has been generated, the corresponding three-dimensional offset for dynamic adjustment in this session is obtained based on the acupoint coordinate adjustment command. Using the offset parameters of the global offset correction model as a priori and the 3D offset of the current session as the local observation, personalized offset model parameters adapted to the current object are obtained through Bayesian posterior update fusion, and the 3D acupoint coordinates are automatically corrected based on the personalized offset model parameters.

[0076] In one possible implementation, the memory 10 may include a program storage area and a data storage area, wherein the program storage area may store the operating system and applications required for at least one function; and the data storage area may store data created during use.

[0077] Furthermore, memory 10 may include read-only memory and random access memory, providing instructions and data to the processor. A portion of the memory may also include NVRAM. The memory stores operating systems and operating instructions, executable modules, or data structures, or subsets thereof, or extended sets thereof, wherein the operating instructions may include various operating instructions for implementing various operations. The operating system may include various system programs for implementing various basic tasks and handling hardware-based tasks.

[0078] Processor 20 can be a central processing unit (CPU), an application-specific integrated circuit, a digital signal processor, a field-programmable gate array, or other programmable logic device. Processor 20 can be a microprocessor or any conventional processor. Processor 20 can call programs stored in memory 10.

[0079] Communication interface 31 can be an interface for the communication module, used to connect with other devices or systems.

[0080] Of course, it should be noted that, Figure 3 The structure shown does not constitute a limitation on the acupoint positioning human-machine intelligent collaborative device in the embodiments of the present invention. In practical applications, the acupoint positioning human-machine intelligent collaborative device may include more than Figure 3 More or fewer components as shown, or combinations of certain components.

[0081] It is understood that if the acupoint positioning human-machine intelligent collaborative method in the above embodiments is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the current technology, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and executes all or part of the steps of the methods in the various embodiments of the present invention. The aforementioned storage medium includes: USB flash drive, mobile hard drive, read-only memory (ROM), random access memory (RAM), electrically erasable programmable ROM, register, hard disk, removable disk, CD-ROM, magnetic disk or optical disk, and other media capable of storing program code.

[0082] Based on this, embodiments of the present invention also provide a medium on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the above-described acupoint positioning human-machine intelligent collaborative method.

[0083] The following describes a computer program product provided by an embodiment of this application. The computer program product described below can be referred to in conjunction with other embodiments described herein.

[0084] A computer program product includes a computer program / instructions that, when executed by a processor, implement the steps of the aforementioned disclosed acupoint positioning human-machine intelligent collaborative method.

[0085] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section.

[0086] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0087] Finally, it should be noted that in this document, relationships such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0088] The above provides a detailed description of the acupoint positioning human-machine intelligent collaborative method, device, equipment, and medium provided by the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A human-machine intelligent collaborative method for acupoint location, characterized in that, include: The three-dimensional acupoint coordinates of the current object are obtained through an artificial intelligence positioning model and output through an interactive visualization interface; When the interactive visualization interface detects that an acupoint coordinate adjustment command has been generated, the corresponding three-dimensional offset for dynamic adjustment in this session is obtained according to the acupoint coordinate adjustment command. Using the offset parameters of the global offset correction model as a priori and the three-dimensional offset of the current session as the local observation, personalized offset model parameters adapted to the current object are obtained through Bayesian posterior update fusion, and the three-dimensional acupoint coordinates are automatically corrected based on the personalized offset model parameters.

2. The acupoint positioning human-machine intelligent collaborative method according to claim 1, characterized in that, When the interactive visualization interface detects that an acupoint coordinate adjustment command has been generated, after obtaining the corresponding three-dimensional offset for dynamic adjustment in this session based on the acupoint coordinate adjustment command, the method further includes: The three-dimensional offset is stored in the session cache corresponding to the current session using a key-value pair structure. The session cache is a temporary memory storage medium and supports selective clearing or persistence after the session is closed. The key in the key-value pair is a unique identifier for the acupoint. The value in the key-value pair is a cumulative three-dimensional offset pattern. The cumulative three-dimensional offset pattern includes at least the three-dimensional offset of a single adjustment and the total three-dimensional offset automatically accumulated after multiple adjustments to the same acupoint. Accordingly, using the offset parameters of the global offset correction model as a priori and the three-dimensional offset of the current session as the local observation, personalized offset model parameters adapted to the current object are obtained through Bayesian posterior update fusion. Based on these personalized offset model parameters, the three-dimensional acupoint coordinates are automatically corrected, including: Calculate statistical characteristic values ​​based on the cumulative three-dimensional offset pattern, including at least one of the average value and proportional relationship; Using the offset parameters of the global offset correction model as priors and the statistical feature values ​​as local observations, personalized offset model parameters adapted to the current object are obtained through Bayesian posterior update fusion, and the coordinates of the three-dimensional acupoints are automatically corrected based on the personalized offset model parameters.

3. The acupoint positioning human-machine intelligent collaborative method according to claim 2, characterized in that, When the interactive visualization interface detects that an acupoint coordinate adjustment command has been generated, after obtaining the corresponding three-dimensional offset for dynamic adjustment in this session based on the acupoint coordinate adjustment command, the method further includes: The personalized offset model parameters are anonymized and differential privacy noise is added to obtain anonymized parameters, which are then uploaded to the server. The server then uses the FedAvg algorithm to aggregate and update the global offset correction model based on the anonymized parameters uploaded by each client, and distributes the updated global offset correction model parameters to each client.

4. The acupoint positioning human-machine intelligent collaborative method according to claim 1, characterized in that, After using the offset parameters of the global offset correction model as a priori, and the 3D offset of the current session as the local observation, personalized offset model parameters adapted to the current object are obtained through Bayesian posterior update fusion. Following the automatic correction of the 3D acupoint coordinates based on these personalized offset model parameters, the process further includes: By statistically analyzing the offset of multiple adjusted acupoints within the same anatomical region, common offset patterns of the anatomical region are obtained, and these common offset patterns are automatically applied to other unadjusted related acupoints within the same anatomical region.

5. The acupoint positioning human-machine intelligent collaborative method according to claim 1, characterized in that, After obtaining the three-dimensional acupoint coordinates of the current object through an artificial intelligence positioning model and outputting them through an interactive visualization interface, the process also includes: The artificial intelligence positioning model outputs nearby candidate points around each acupoint. The neighboring candidate points are clustered to obtain clusters; Accordingly, the three-dimensional acupoint coordinates of the current object are obtained through an artificial intelligence positioning model and output through an interactive visualization interface, including: The three-dimensional acupoint coordinates and the coordinates of the center point corresponding to the cluster are output through an interactive visualization interface.

6. The acupoint positioning human-machine intelligent collaborative method according to claim 1, characterized in that, The three-dimensional acupoint coordinates of the current object are obtained through an artificial intelligence positioning model and output through an interactive visualization interface, including: The 3D point cloud of the current object is acquired using an RGB-D camera, and based on the 3D point cloud, 3D feature points are acquired using a 3D feature point detection model. Interpolation points are obtained by performing interpolation processing based on the feature points of the three-dimensional human body parts. Based on the interpolation points and the three-dimensional feature points, the coordinates of the three-dimensional acupoints are determined using the adaptive multi-segment bone measurement method and meridian database mapping; the meridian database includes a standardized acupoint location table.

7. The acupoint positioning human-machine intelligent collaborative method according to claim 1, characterized in that, The three-dimensional acupoint coordinates of the current object are obtained through an artificial intelligence positioning model and output through an interactive visualization interface, including: Based on the three-dimensional acupoint coordinates, the locations are marked on the color map of the interactive visualization interface, and the acupoint location reliability and suggested acupoint depth output by the artificial intelligence positioning model are overlaid and displayed.

8. A human-machine intelligent collaborative device for acupoint positioning, characterized in that, include: The interface output module is used to obtain the three-dimensional acupoint coordinates of the current object through an artificial intelligence positioning model and output them through an interactive visual interface. The interaction module is used to obtain the corresponding three-dimensional offset for dynamic adjustment in this session based on the acupoint coordinate adjustment command when the interactive visualization interface generates an acupoint coordinate adjustment command. The correction module is used to obtain personalized offset model parameters adapted to the current object by using the offset parameters of the global offset correction model as a priori and the three-dimensional offset of the current session as the local observation value, through Bayesian posterior update fusion, and automatically correct the three-dimensional acupoint coordinates based on the personalized offset model parameters.

9. A human-machine intelligent collaborative device for acupoint positioning, characterized in that, include: Memory, used to store computer programs; A processor is configured to implement the acupoint positioning human-machine intelligent collaborative method as described in any one of claims 1 to 7 when executing the computer program.

10. A medium, characterized in that, The medium stores computer-executable instructions, which, when loaded and executed by a processor, implement the acupoint positioning human-machine intelligent collaborative method as described in any one of claims 1 to 7.

Images