Intelligent electric moxibustion physiotherapy control method and system based on acupoint recognition and medium

Abstract

The invention relates to an intelligent electric moxibustion physiotherapy control method and system based on acupoint recognition and a medium, and belongs to the technical field of traditional Chinese medicine physiotherapy, the control method comprises the steps that image information of a target object is collected, and the image information comprises a first image used for identity recognition and a second image used for acupoint recognition; determining identity information of the target object based on the first image; acquiring an electric moxibustion physiotherapy scheme corresponding to the target object from a preset database according to the identity information; performing acupoint key point detection based on the second image, and outputting an acupoint recognition result; guiding the target object to be in place based on the electric moxibustion physiotherapy scheme, screening out a target acupoint coordinate in the electric moxibustion physiotherapy scheme from the acupoint identification result, and performing light spot projection; and controlling the electric pulse generator to output a multi-channel electric pulse signal to the target acupoint. The acupoint positioning accuracy and the electric moxibustion physiotherapy efficiency can be improved.

Description

Technical Field

[0001] This application relates to the field of traditional Chinese medicine physiotherapy technology, and in particular to an intelligent electro-moxibustion physiotherapy control method, system and medium based on acupoint recognition. Background Technology

[0003] Traditional acupuncture primarily involves manual acupuncture. However, training qualified acupuncturists is extremely expensive and time-consuming. Furthermore, individual acupuncture treatments are lengthy and labor-intensive, and individual differences among patients can lead to acupuncturists becoming fatigued and making diagnostic errors. Therefore, manual acupuncture treatment suffers from numerous drawbacks, including low accuracy, low efficiency, and high cost. Thus, with the advent of the artificial intelligence era, automating the acupuncture treatment process and minimizing human labor has become a significant challenge and opportunity.

[0004] Currently, in the process of automating acupuncture and physiotherapy, how to achieve automatic acupoint identification and positioning is a key issue that needs to be considered. In practical applications, because different patients have different treatment plans tailored to them, conventional automatic acupoint identification methods usually do not take into account the individual differences of different patients. They are easily affected by factors such as operator experience, patient body shape differences (such as obesity or thinness), and postural deviations, which pose a risk of positioning deviation. This can cause the electrode to be attached to a location that deviates from the actual acupoint, thereby reducing the accuracy of acupoint positioning and the efficiency of electro-moxibustion physiotherapy. Summary of the Invention

[0005] To improve the accuracy of acupoint location and the efficacy of electro-moxibustion therapy, this application provides an intelligent electro-moxibustion therapy control method, system, and medium based on acupoint recognition.

[0006] Firstly, this application provides an intelligent electro-moxibustion therapy control method based on acupoint recognition, employing the following technical solution: A smart electro-moxibustion therapy control method based on acupoint recognition, the control method comprising: Image information of the target object is acquired, including a first image for identification and a second image for acupoint identification; The identity information of the target object is determined based on the first image; Based on the identity information, the electro-moxibustion therapy plan corresponding to the target object is obtained from a preset database; Based on the second image, key points of acupoints are detected, and acupoint recognition results are output. Based on the electro-moxibustion therapy plan, the target object is positioned, and the coordinates of the target acupoints in the electro-moxibustion therapy plan are selected from the acupoint recognition results and light spots are projected. The control pulse generator outputs multi-channel electrical pulse signals to the target acupoint.

[0007] Secondly, this application provides an intelligent electro-moxibustion therapy control system based on acupoint recognition, which adopts the following technical solution: A smart electro-moxibustion therapy control system based on acupoint recognition, the control system comprising: The image acquisition module is used to acquire image information of the target object, the image information including a first image for identification and a second image for acupoint identification; The identity recognition module is used to determine the identity information of the target object based on the first image; An electro-moxibustion therapy plan acquisition module is used to retrieve the electro-moxibustion therapy plan corresponding to the target object from a preset database based on the identity information. The acupoint recognition module is used to detect key acupoints based on the second image and output the acupoint recognition results. The positioning instruction module is used to guide the target object into position based on the electro-moxibustion therapy plan; The projection control module is used to filter out the coordinates of the target acupoints in the electro-moxibustion therapy plan from the acupoint recognition results and project light spots. The electro-moxibustion control module is used to control the electric pulse generator to output multi-channel electric pulse signals to the target acupoint.

[0008] Thirdly, this application provides a computer-readable storage medium, which adopts the following technical solution: A computer-readable storage medium storing a computer program that can be loaded by a processor and executed as in any of the methods in the first aspect.

[0009] In summary, this application includes at least one of the following beneficial technical effects: Based on the combination of image acquisition and deep learning models, it achieves full automation from identity recognition to acupoint recognition, reducing reliance on professional TCM doctors; through light spot projection and voice guidance technology, it enhances the visualization and interactivity of the treatment process, improving patient compliance and experience; through multi-channel electrical pulse output and automatic control technology, it achieves precise regulation of treatment parameters, ensuring the safety and effectiveness of treatment. This technical solution not only promotes the modernization of TCM physiotherapy techniques but also provides an efficient and intelligent solution for chronic disease rehabilitation in the context of an aging society. Attached Figure Description

[0010] Figure 1 This is a schematic diagram of the first process of an intelligent electro-moxibustion therapy control method according to one embodiment of this application.

[0011] Figure 2 This is a second flowchart of an intelligent electro-moxibustion therapy control method according to one embodiment of this application.

[0012] Figure 3 This is a schematic diagram of the third process of an intelligent electro-moxibustion therapy control method according to one embodiment of this application.

[0013] Figure 4 This is a schematic diagram of the fourth process of the intelligent electro-moxibustion physiotherapy control method according to one embodiment of this application.

[0014] Figure 5 This is a schematic diagram of the fifth process of the intelligent electro-moxibustion physiotherapy control method according to one embodiment of this application.

[0015] Figure 6 This is a schematic diagram of the sixth process of the intelligent electro-moxibustion physiotherapy control method according to one embodiment of this application.

[0016] Figure 7 This is a schematic diagram of the seventh process of the intelligent electro-moxibustion physiotherapy control method according to one embodiment of this application.

[0017] Figure 8 This is a schematic diagram of the eighth process of the intelligent electro-moxibustion physiotherapy control method according to one embodiment of this application.

[0018] Figure 9 This is a schematic diagram illustrating the use scenario of the intelligent electro-moxibustion physiotherapy control system according to one embodiment of this application. Detailed Implementation

[0019] To make the purpose, technical solution, and advantages of this application clearer, the following description is provided in conjunction with the appendix. Figure 1 -Appendix Figure 9 The present application will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the application.

[0020] This application discloses an intelligent electro-moxibustion therapy control method based on acupoint recognition.

[0021] Reference Figure 1 A smart electro-moxibustion therapy control method based on acupoint recognition, the control method includes: Step S101: Acquire image information of the target object; the image information includes a first image for identification and a second image for acupoint identification; In this embodiment, a camera can be mounted above a standardized hospital bed where the target person (i.e., a physiotherapy patient) is lying, and full-body images of the patient can be captured through continuous shooting, including a first image for identification and a second image for acupoint identification. The first image can be a facial image for individual identification, and its acquisition requires a clear image, a clean background, and no obstructions. The second image can be a body image containing acupoints, covering the target person's torso, limbs, and other major meridian areas to ensure the complete presentation of key acupoints.

[0022] Step S102: Determine the identity information of the target object based on the first image; In this embodiment of the application, the system uses a RetinaFace-based face detection model to locate the face region in the first image, extracts the facial region image, and then uses an ArcFace face recognition model to extract and compare features.

[0023] Specifically, RetinaFace is a face detection method based on the RetinaNet architecture. By introducing a multi-task learning mechanism, it can simultaneously predict face bounding boxes, key point locations, and facial attributes, exhibiting high detection accuracy and robustness. ArcFace, on the other hand, is an identity recognition model based on a deep convolutional neural network. By introducing the ArcCos Loss function, it effectively improves the discriminative ability of facial features, maximizing the feature distance between different individuals and minimizing the feature distance between the same individual. Through this model, the system can accurately identify the identity ID of the target object, which serves as the basis for retrieving subsequent treatment plans. This process not only enhances the system's personalized service capabilities but also provides a foundation for data management and tracking during subsequent treatment.

[0024] Step S103: Obtain the electro-moxibustion therapy plan corresponding to the target object from the preset database based on the identity information; The pre-set database is a structured patient information management system that stores each patient's basic information (such as age, gender, medical history, etc.) and a personalized electro-moxibustion therapy plan corresponding to their health condition. The electro-moxibustion therapy plan is generally formulated by a professional TCM physician based on the patient's condition, constitution, meridian status, and other factors. The plan may include a list of acupoints to be stimulated, a sequence of body positions, stimulation intensity (such as electrical pulse frequency and amplitude), duration, stimulation order, etc.

[0025] The system uses the user's ID as an index to quickly retrieve and load the corresponding treatment plan, ensuring the accuracy and personalization of subsequent operations. This process demonstrates the system's advantages in personalized medicine and also reflects the integrated application of artificial intelligence in traditional Chinese medicine diagnosis and treatment.

[0026] Step S104: Detect key points of acupoints based on the second image and output the acupoint recognition results; In this embodiment, the system uses PIPNet (Pixel-in-Pixel Network) as the main acupoint recognition model, which uses ResNet50 as the backbone network for feature extraction and combines the Neighbor Regression Module (NRM) for key point coordinate prediction.

[0027] Specifically, unlike traditional heatmap regression methods, PIPNet employs a pixel-level prediction-based regression strategy. It directly predicts the coordinate offsets of key points on low-resolution feature maps, avoiding the computational complexity and information loss caused by upsampling in traditional methods. The NRM module further improves recognition accuracy by introducing geometric relationships between adjacent key points, especially in cases of densely packed acupoints with similar shapes, effectively distinguishing adjacent acupoints and improving localization accuracy. During the training phase, the model uses a transfer learning strategy, fine-tuning it with an annotated dataset of traditional Chinese medicine acupoint images to adapt to acupoint recognition needs under different body types and postures. The final output acupoint recognition results are a series of acupoints with spatial coordinates, providing precise physical location information for subsequent light projection and electrode patching.

[0028] Step S105: Based on the electro-moxibustion therapy plan, guide the target object to position, select the coordinates of the target acupoints in the electro-moxibustion therapy plan from the acupoint recognition results, and project light spots. Specifically, the system determines the body position (such as supine, prone, left lateral, right lateral, etc.) that the target subject needs to adopt based on the distribution of acupoints involved in the electro-moxibustion therapy plan, and guides the target subject to complete the body position adjustment through voice prompts.

[0029] Furthermore, after the target object is positioned, the system uses a body position determination network (typically a convolutional neural network based on an SSD architecture) to identify the current body position, ensuring it matches the preset position in the physiotherapy plan. If the position is incorrect, the system will continue to provide voice prompts until the target object is properly adjusted. After confirming the position is correct, the system filters acupoint coordinates related to the current physiotherapy plan from the acupoint recognition results and maps them to the actual physical space using a spatial transformation algorithm to determine the coordinates of the light spot projection. Specifically, the system calculates the translation and rotation parameter matrices based on the known correspondence between acupoint recognition coordinates and light spot projection coordinates, thereby achieving coordinate system transformation.

[0030] Step S106: Control the electrical pulse generator to output multi-channel electrical pulse signals to the target acupoint.

[0031] The electrical pulse generator is typically composed of a multi-function generator, capable of outputting electrical stimulation signals of different frequencies, waveforms (such as square waves, triangular waves, and pulse waves), and amplitudes. Based on the parameters set in the electro-moxibustion therapy plan, the system sends commands to the electrical pulse generator through the control module to adjust the characteristics of the output signal.

[0032] For example, the system can be set with different frequency ranges (e.g., 2–100 Hz) and pulse widths (e.g., 100–500 μs) for different conditions to achieve different therapeutic effects. Simultaneously, the system also features automatic connection detection. After the caregiver attaches the electrode pads to the light projection position and connects the output cable, the system detects the circuit's on / off state and impedance changes to determine if the electrodes are correctly connected, ensuring the safety and effectiveness of the treatment. The electrical pulse signals act on acupoints through the electrode pads, stimulating meridians, regulating qi and blood, and improving local microcirculation to achieve therapeutic effects such as pain relief and promoting recovery.

[0033] In the above implementation, the combination of image acquisition and deep learning models automates the entire process from identity recognition to acupoint identification, reducing reliance on professional TCM doctors. Light projection and voice guidance technologies enhance the visualization and interactivity of the treatment process, improving patient compliance and experience. Multi-channel electrical pulse output and automatic control technologies enable precise regulation of treatment parameters, ensuring the safety and effectiveness of the treatment. This technical solution not only promotes the modernization of TCM physiotherapy techniques but also provides an efficient and intelligent solution for chronic disease rehabilitation in an aging society.

[0034] Reference Figure 2 As a further implementation of the intelligent electro-moxibustion therapy control method, after the step of acquiring the image of the target object, the method further includes: Step S201: Convert the image of the target object into a grayscale image; In this embodiment of the application, the conversion formula is: I gray = 0.2989×R + 0.5870×G + 0.1140×B; In the above formula, R, G, and B represent the red, green, and blue channels of the original color image, respectively; I gray This is the converted grayscale image.

[0035] Step S202: Normalize the grayscale image to obtain a normalized image; Specifically, the grayscale image above is normalized to [0, 255], and the calculation formula is as follows: In the above formula, I max and I minThe grayscale images I are respectively gray The maximum and minimum gray values ​​in I n This is the normalized image output.

[0036] Step S203: Adjust the contrast of the normalized image based on the grayscale histogram equalization algorithm to obtain the preprocessed image, including the preprocessed first image and the second image.

[0037] Specifically, the formula for the gray-level histogram equalization algorithm is expressed as follows: In the above formula, H(i) represents the pixel value after histogram equalization, round represents rounding the pixel value H(i), P(j) represents the frequency of the pixel with gray value j in the image, CDF(i) is the cumulative distribution function value of gray value i, CDF(min) is the cumulative distribution function value of the minimum gray value, M and N represent the image width and height respectively, L represents the range of pixel values, and min represents the minimum pixel value of the image before histogram equalization.

[0038] In the above embodiments, since the original image acquired is a color image, grayscale preprocessing is required. The first and second images after preprocessing are used for identity and acupoint recognition, respectively, thereby improving the accuracy and efficiency of image processing.

[0039] Reference Figure 3 As an implementation of steps S102 and S103, the step of determining the identity information of the target object based on the first image and obtaining the electro-moxibustion therapy plan corresponding to the target object from the pre-stored database according to the identity information includes: step S301, inputting the first image into a pre-trained face detection model for detection, and cropping to obtain a face region image; As one implementation method for a face detection model, this application uses the RetinaFace face detection framework to implement a deep learning-based face detection model. RetinaFace is a face detection framework based on RetinaNet, which achieves efficient and accurate face detection mainly through multi-task learning and lightweight network design.

[0040] In this embodiment, the RetinaFace face detection framework, by introducing a multi-task learning mechanism, can not only predict face bounding boxes, but also simultaneously predict facial key points (such as eyes, nose, and mouth) and facial attributes (such as pose and occlusion). This allows it to maintain high face detection accuracy even under complex backgrounds, varying lighting conditions, and different angles. The model's lightweight design enables efficient operation on embedded devices or edge computing platforms, meeting the deployment requirements of this application in real-world medical scenarios. Through the processing of the RetinaFace model, the system can accurately extract the face region image from the first image, providing high-quality input data for subsequent identity recognition.

[0041] Step S302: Input the face region image into a pre-trained face recognition model for recognition, and output the identity ID as the identity information of the target object; After passing through the face detection model, the face image corresponding to the face bounding box can be used as input to identify the patient's identity ID using a deep learning-based face detection model.

[0042] In this embodiment, the ArcFace face recognition model can be used to identify the patient and determine their identity ID. This model is an identity recognition algorithm based on a deep convolutional neural network. Its core lies in the introduction of the ArcCos Loss function, which significantly improves the model's discriminative ability by maximizing inter-class distance and minimizing intra-class distance in the feature space. Compared with traditional Softmax loss functions or SphereFace models, ArcFace exhibits higher recognition accuracy and stronger generalization ability on large-scale face datasets.

[0043] In the application scenario of this application, the system encodes the input facial image into a high-dimensional feature vector using the ArcFace model, compares it with the feature vectors registered in the database, and finally outputs the identity ID that best matches it. This process not only enables rapid identification of the target object, but also provides a basis for subsequent retrieval of personalized physical therapy plans.

[0044] Step S303: Search the preset database based on the identity ID to obtain the electro-moxibustion therapy plan corresponding to the identity ID.

[0045] After obtaining the current patient's identity ID, a search request can be sent to a server containing a preset database based on that ID. This retrieves the electro-moxibustion therapy plan corresponding to the identity ID, found by the server based on the search request. Specifically, the preset database stores all patients' personal information and corresponding electro-moxibustion therapy plans.

[0046] Understandably, electroacupuncture therapy plans are personalized treatment plans developed by professional TCM doctors based on the individual differences of patients. These plans typically include: a list of acupoints to be stimulated, electrode channel allocation strategies, electrical pulse parameters (such as frequency, amplitude, and pulse width), therapy time arrangements, and body position sequences. After obtaining the patient's identity ID, the system sends a retrieval request to the server via a network communication interface. Upon receiving the request, the server uses database query statements (such as SQL) to find the corresponding therapy plan based on the identity ID and returns the result to the control system. This process ensures the personalization and precision of subsequent therapy operations and is a key technological support for realizing the concept of "treatment tailored to the individual."

[0047] The above implementation achieves structured management of patient information and precise retrieval of physiotherapy plans, giving the entire system a high degree of intelligence and automation. This process not only improves the efficiency and safety of TCM electro-moxibustion physiotherapy, but also provides technical support for new medical models such as smart healthcare and remote physiotherapy, possessing significant clinical application value and promising prospects for promotion.

[0048] Reference Figure 4 As one implementation of step S104, the step of detecting key acupoints based on the second image and outputting acupoint recognition results includes: Step S401: Input the second image into the ResNet50 backbone network to extract image features; In this embodiment of the application, PIPNet (Pixel-in-Pixel Net) key point detection neural network can be used for automatic acupoint identification and positioning. This key point detection neural network uses ResNet50 as the network backbone to extract image features from the input whole body image.

[0049] Specifically, ResNet50 is a classic residual network whose core lies in the introduction of a "residual block" structure. By using skip connections, it solves the gradient vanishing problem in deep networks, allowing the network to stack to more than 50 layers without sacrificing training efficiency, thereby extracting deeper semantic features from images. In this application, ResNet50 serves as the "backbone network" of the keypoint detection network. Its input is a second image (usually a full-body image of the target object), and its output is a low-resolution, high-dimensional feature map. This feature map preserves crucial structural information in the image, such as human contours, limb positions, and joint distribution, which are essential for subsequent acupoint localization. The introduction of ResNet50 not only improves the robustness of feature extraction but also provides a high-quality semantic foundation for subsequent coordinate regression tasks.

[0050] It should be noted that the keypoint detection neural network in this application differs from traditional heatmap regression methods (such as Hourglass, U-Net, etc.). It abandons the upsampling process and directly predicts keypoint coordinates on low-resolution feature maps, thereby significantly improving the model's computational efficiency while maintaining recognition accuracy. The core idea of ​​PIPNet is to predict two key pieces of information at each pixel—a "score" and an "offset." The score is used to determine whether the pixel is a candidate location for a certain acupoint, while the offset is used to correct the spatial deviation between the pixel and the actual acupoint coordinates. This strategy is called PIP regression (Pixel-in-Pixel Regression), and its advantage lies in achieving high-precision keypoint localization without relying on high-resolution images, making it particularly suitable for the complex backgrounds and small-scale acupoint targets in the full-body images of this application.

[0051] Step S402: The image features are regressed by the nearest neighbor regression module to output the two-dimensional coordinate set of all acupoints and obtain the acupoint recognition result; wherein, the nearest neighbor regression module is used to predict the main offset of each acupoint and the associated offset of adjacent acupoints.

[0052] Furthermore, this application introduces a Neighbor Regression Module (NRM) based on PIPNet to improve the accuracy and robustness of acupoint recognition. The core concept of the NRM module is to predict not only the offset of each acupoint itself, but also the associated offset of its neighboring acupoints. Since the distribution of acupoints in the human body has a clear spatial structure, there are certain geometric relationships between adjacent acupoints. For example, "Zusanli" and "Yanglingquan" are linearly arranged on the leg, and "Hegu" and "Quchi" have a certain angular relationship on the arm. The NRM module utilizes this spatial correlation, enhancing the model's ability to discriminate acupoint locations by introducing the offset prediction of neighboring acupoints. Especially in cases where acupoints are densely packed and have similar shapes, it can effectively avoid misidentification and misplacement problems.

[0053] Specifically, the NRM module introduces multiple "neighbor offsets" (e.g., 3 or 5) into the prediction of each keypoint and incorporates them into the optimization process of the loss function, thereby enabling the model to learn richer spatial context information during training. This multi-task learning mechanism not only improves the model's generalization ability but also enhances its adaptability to different body sizes and poses.

[0054] Furthermore, this keypoint detection neural network requires training before use. The training dataset is pre-collected and labeled by professional TCM doctors. This is particularly crucial because acupoint recognition differs from general image recognition tasks, requiring highly specialized and standardized labeling. Training data is typically generated by TCM doctors using infrared markers, laser positioning, and other methods to determine the precise coordinates of each acupoint on standard human models or volunteers, and then labeling these coordinates in the collected full-body images. During training, the system employs a transfer learning strategy. First, PIPNet is pre-trained on a large-scale general keypoint dataset (such as COCO, MPII, etc.) to give it basic human keypoint detection capabilities. Then, fine-tuning is performed on a TCM acupoint dataset to adapt the model to the specific distribution patterns of TCM acupoints. This training strategy not only improves the model's convergence speed but also effectively alleviates the problem of limited TCM acupoint data, ensuring the model's stability and accuracy in practical applications.

[0055] In the above embodiments, the system can accurately identify the two-dimensional coordinates of multiple acupoints against complex backgrounds, providing solid technical support for subsequent operations such as light spot projection, body positioning guidance, and electrical pulse stimulation. This technical solution provides a practical and feasible technical path for achieving standardization, personalization, and automation of TCM diagnosis and treatment, and has significant clinical application value and social promotion significance.

[0056] As one implementation method of electroacupuncture therapy, it includes an acupoint list, a channel allocation strategy, and a body position sequence. The acupoint list refers to a set of specific acupoints requiring electrical pulse stimulation, formulated by professional physicians based on Traditional Chinese Medicine theory and individual patient differences, such as "Zusanli," "Hegu," and "Yanglingquan." The channel allocation strategy refers to assigning a corresponding electrode channel number and its stimulation parameters (such as frequency, amplitude, pulse width, etc.) to each acupoint in a multi-channel electrical pulse output system, ensuring that multiple acupoints can be stimulated synchronously or asynchronously without interference. The body position sequence refers to the order in which the patient needs to adjust their body position during a complete therapy session, such as "supine → prone → left lateral → right lateral." This sequence is set based on the operability of the acupoints and the convenience of electrode attachment. These three elements constitute a structured and executable set of therapy instructions, serving as the basis for subsequent automatic control.

[0057] Reference Figure 5 As one implementation of steps S105 and S106, the steps of guiding the target object to position based on the electro-moxibustion therapy plan, selecting the coordinates of the target acupoints in the electro-moxibustion therapy plan from the acupoint identification results and projecting light spots, and controlling the electric pulse generator to output multi-channel electric pulse signals to the target acupoints include: Step S501: Generate position adjustment prompts sequentially based on the position sequence to guide the target object to adjust to the target position in sequence; Specifically, based on the position sequence in the electro-moxibustion therapy plan and the actual position of the target subject (obtained by the positioning judgment module), the system generates corresponding prompts through the speech synthesis module, such as "Please adjust to a supine position" or "Please lie on your right side." This prompt not only improves patient compliance but also ensures the accuracy of subsequent acupoint recognition and electrode attachment. Position recognition itself relies on a position judgment network model (such as a classification model based on the SSD architecture). This model analyzes the target subject's full-body image, extracting features such as limb posture and spinal alignment to determine if the current position conforms to preset standards. If the recognition result does not match the target position, the system will continue to provide prompts until the target subject is correctly positioned.

[0058] Step S502: Based on the target object's current body position and the acupoint list, filter the target acupoint coordinate set from the acupoint recognition results; wherein, the current body position includes supine, prone, left lateral, or right lateral. This step reflects the spatial adaptability principle of traditional Chinese medicine physiotherapy, meaning that the degree of acupoint exposure and the feasibility of electrode adhesion vary depending on the body position. For example, the "Mingmen" acupoint is located in the middle of the spine and is difficult to operate in a supine position, but is easier to locate in a prone position; the "Huantiao" acupoint is located on the buttocks and is easier to reach in a side-lying position. Therefore, the system selects only the acupoints that can be operated in the current body position as the target acupoints in each stage, avoiding misoperation or ineffective stimulation caused by body position limitations.

[0059] Specifically, the system matches the list of acupoints with the coordinate information in the acupoint recognition results, and combines this with the accessibility judgment of acupoints from the current body position to filter out a set of acupoint coordinates suitable for the current stage of operation. This process typically relies on a spatial accessibility model and a body position-acupoint mapping table. The former judges the operability of a certain acupoint in a specific body position based on human anatomy knowledge, while the latter is formulated by traditional Chinese medicine experts to clarify the list of acupoints that should be prioritized for operation in different body positions.

[0060] Step S503: Perform spatial mapping calculation based on the target acupoint coordinate set to generate a light point projection coordinate set; Since the acupoint coordinates obtained from image recognition are pixel coordinates, and the actual projection onto the human body surface requires consideration of the mapping relationship between physical space and image space, spatial transformation processing is necessary. Common spatial transformation methods include affine transformation and perspective transformation. The core of these methods is to calculate a transformation matrix using known reference point pairs (such as image coordinates and actual physical coordinates), thereby converting pixel coordinates into physical space coordinates.

[0061] For example, the system can establish a correspondence between image space and physical space using a calibration board, solve the transformation matrix using the least squares method, and then convert the pixel coordinates of acupoints into the physical coordinates required by the projection device. This coordinate set is the light spot projection coordinate set, which is used to control the projection device to accurately project the light spot onto the target acupoint location.

[0062] Step S504: Generate a black background image based on the target body position, and light up the corresponding pixels on the black background image according to the light spot projection coordinate set to form a projection image with light spots. Specifically, the system generates a black background image (i.e., a completely black image) that matches the patient's body surface area based on the current position. Then, the pixel values ​​at the corresponding projection coordinates of the light points on the image are set to the maximum brightness value (e.g., 255), thus forming visual light point markers. This image generation method not only facilitates subsequent projection operations but also provides caregivers with intuitive visual guidance, enabling them to quickly identify and attach electrode pads. For example, in the "prone" position, the system illuminates light points corresponding to acupoints such as "Mingmen" and "Shenshu" on the back of the image; while in the "supine" position, it illuminates light points for acupoints on the front, such as "Zusanli" and "Guanyuan."

[0063] Step S505: Control the light spot projection device to project the image onto the surface of the target object; The projection equipment is typically a high-precision laser projector or a DLP projection system, and its projection accuracy and image resolution directly affect the accuracy of the light spot positioning. The system ensures that the light spot accurately lands on the target acupoint by controlling the projector's projection angle, focal length, and coordinate mapping relationship.

[0064] To improve positioning accuracy, the system may employ a dynamic correction mechanism. This involves using an infrared camera or depth camera to detect the patient's surface contours before each projection and correcting the projection coordinates in real time to compensate for errors caused by slight patient movements or breathing. This real-time feedback mechanism significantly enhances the stability and reliability of the light spot projection.

[0065] It should be noted that the light spots projected onto the acupoints on the target patient's body can serve as visual cues for caregivers, providing a basis for subsequent wiring operations. Caregivers can visually observe the light spots projected onto the patient's acupoints and then operate the electrical pulse generator to perform electrical pulse stimulation therapy.

[0066] Step S506: Based on the channel allocation strategy, control the electrical pulse generator to output multi-channel electrical pulse signals to the target acupoint.

[0067] The electrical pulse generator is typically a multi-channel function generator, capable of independently controlling the waveform type (e.g., square wave, triangle wave, sine wave), frequency (e.g., 2Hz–100Hz), amplitude (e.g., 0–50mA), and duration (e.g., 10s–60s) of each output channel. Based on a channel allocation strategy, the system binds each target acupoint to its corresponding electrode channel and sends control commands via a communication interface (e.g., RS485, CAN, USB) to ensure that each acupoint receives electrical stimulation according to preset parameters.

[0068] For example, the system can be set to simultaneously provide 10Hz, 20mA square wave stimulation to "Zusanli" and "Yanglingquan" acupoints for 30 seconds; while "Hegu" and "Quchi" acupoints can receive 50Hz, 15mA triangular wave stimulation for 20 seconds. This multi-channel, multi-parameter control method not only improves the flexibility of physiotherapy but also enhances its adaptability to different conditions.

[0069] The above implementation achieves a complete closed-loop control from image recognition to physical therapy. This process not only improves the intelligence and standardization of TCM electro-moxibustion therapy, but also significantly reduces the reliance on professional acupuncturists. While ensuring the accuracy, safety and personalization of the therapy, it also lays the foundation for the popularization and promotion of intelligent therapy equipment in the future.

[0070] Reference Figure 6 As a further implementation of the intelligent electro-moxibustion therapy control method, after the step of generating position adjustment prompts sequentially based on the position sequence to guide the target subject to adjust to the target position in sequence, it also includes: Step S601: Acquire real-time body position images of the target object; The image acquisition module typically uses a high-resolution industrial camera to capture full-body images of the target object from a top-down or side-view angle. The image content should completely cover key parts of the human body, such as the torso, limbs, and head, to facilitate the identification of its posture features. Since this image is used for body position determination, high requirements are placed on image clarity, illumination uniformity, and background interference control. Image enhancement processing may also be necessary to improve the accuracy of subsequent recognition.

[0071] Step S602: Based on the real-time body posture image, identify the body posture features of the target object and determine the current body posture state. Specifically, this process relies on a body posture judgment network model, which is essentially a deep learning-based image classification or posture estimation model used to extract structured posture information of the human body from the image and determine its body posture category.

[0072] In this embodiment, a body position determination network based on the SSD (Single Shot MultiBox Detector) architecture can be used. This model detects human contours, limb distribution, and spinal alignment through multi-scale feature maps, extracts key posture features, and then determines whether the current body position is "supine," "prone," "left lateral," or "right lateral." The training dataset of the model consists of a large number of labeled body position images, covering samples of different body types, ages, and clothing conditions, in order to improve the model's generalization ability.

[0073] Step S603: Determine whether the current body position matches the target body position in the body position sequence; if not, proceed to step S604; if yes, continue to the next step. Step S604: Output posture correction prompts.

[0074] The process involves comparing the identified "current body position" with the "preset target body position in the electro-moxibustion therapy plan" to determine if they match. If they match, the target body is considered correctly positioned, and subsequent acupoint identification and light spot projection operations can continue. If they do not match, a body position correction mechanism is triggered to ensure the patient is in the correct operating position. This process relies on comparing the current body position label with the target body position label using strings or values, and the flow of the process is determined based on the comparison result. For example, in a certain electro-moxibustion therapy session, the current target body position should be "supine," but the system identifies the current body position as "prone," which is considered a mismatch, triggering subsequent correction prompts.

[0075] Furthermore, if the judgment result is "the current body position does not conform to the target body position," the system enters the body position correction prompt stage. This step is crucial for achieving human-computer interaction and behavioral guidance. Its purpose is to provide clear body position adjustment guidance to the target subject through voice, images, text, and other means, helping them complete the correct body position transition. For example, the system can issue a prompt such as "Please adjust to a supine position" through the voice synthesis module, or display the corresponding body position diagram on the display module. This prompt mechanism not only improves patient compliance but also effectively reduces the risk of acupoint recognition failure or electrode adhesion error caused by incorrect body position. In addition, the system can also combine historical body position recognition results to analyze whether the target subject has a tendency to repeatedly deviate from the correct body position, and adjust the prompt frequency or prompt method accordingly to achieve a personalized human-computer interaction experience.

[0076] It should be noted that the construction and training of the posture judgment network model is one of the core challenges in the entire posture recognition and judgment process. This model needs to possess good posture robustness, maintaining high recognition accuracy under complex environments with varying lighting, occlusion, and background interference. To this end, the system typically employs a transfer learning strategy, first pre-training the model on large-scale general posture datasets (such as COCO and MPII), and then fine-tuning it on specific posture datasets.

[0077] Furthermore, to further enhance the model's generalization ability, data augmentation techniques (such as rotation, scaling, and brightness adjustment) can be introduced during training to simulate body position images in various real-world scenarios. Taking the body position judgment network based on the SSD architecture in this embodiment as an example, its structure includes a feature extraction backbone network, a multi-scale feature map generation module, and a classification and regression head, which can simultaneously complete human body detection and body position classification tasks, and is suitable for the real-time recognition requirements of body position status in this application.

[0078] The above implementation not only ensures that the target subject is in the correct position before electroacupuncture therapy, but also provides a reliable spatial basis for subsequent acupoint identification and electrode attachment. Through the deep integration of deep learning, image processing, and human-computer interaction technologies, the system significantly reduces reliance on professional medical personnel while improving the accuracy, safety, and personalization of the therapy.

[0079] Reference Figure 7 As one implementation of step S503, the step of generating a light spot projection coordinate set by performing spatial mapping calculation based on the target acupoint coordinate set includes: Step S701: Calculate the translation and rotation parameters based on the preset projection coordinate sample set and the target acupoint coordinate set. The calculation formula is: (Δx, Δy, Δθ) T =(x d ,y d ,θ d ) T -(x D ,y D ,θ D ) T Where (Δx, Δy, Δθ) represent translation and rotation parameters, (x... d ,y d ,θ d (x) represents the preset projection coordinate sample. D ,y D ,θ D () represents the coordinates of the target acupoint, and T represents the matrix transpose symbol; Step S702: Calculate the correction amount for the coordinates to be projected based on the translation and rotation parameters. The calculation formula is as follows: Among them, (x i,y i ,θ i () represents the coordinates to be projected. This represents the correction amount for the coordinates to be projected; The process involves mapping the coordinates of all acupoints to be projected in the image space to the physical space using a transformation matrix, and calculating their offset relative to the original image coordinates. This offset is the "position correction value" that each acupoint should be projected into in the physical space, ensuring that the light spot accurately lands on the target acupoint. For example, if the coordinates of an acupoint in the image space are (512, 384), and its correction value calculated by the transformation matrix is ​​(12.5, -8.3), then its projected coordinates in the physical space should be (524.5, 375.7).

[0080] Step S703: Correct the target acupoint coordinate set based on the correction amount, and output the light spot projection coordinate set.

[0081] The light spot projection coordinate set is used to control the projection position of the light spots on the projection device, ensuring that each light spot accurately lands on the target acupoint. It is worth noting that this process applies not only to acupoint projection in a single body position but also to dynamic coordinate adjustments under multiple body position changes. For example, when the patient changes from a supine to a prone position, the system will re-acquire images, identify new acupoint coordinates, and repeat the above spatial mapping process to ensure the accuracy of light spot projection under different body positions.

[0082] In the above implementation, an affine transformation model between image space and physical space is constructed. Combined with matrix operations and coordinate correction mechanisms, a precise mapping from acupoint recognition results to the coordinates of the projected light spot is achieved. This process not only solves the spatial misalignment problem between image recognition and physical execution but also ensures the stability and accuracy of the light spot projection under different body positions, patients, and equipment configurations. By introducing mathematical modeling and spatial transformation algorithms, the system achieves high-precision conversion of acupoint coordinates, providing reliable visual guidance for subsequent electrode adhesion and electrical pulse stimulation, significantly improving the automation level and clinical applicability of the intelligent electro-moxibustion therapy system.

[0083] Reference Figure 8 As a further implementation of the intelligent electro-moxibustion therapy control method, before the step of controlling the output of multi-channel electrical pulse signals from the electrical pulse generator to the target acupoint based on the channel allocation strategy, the method further includes: Step S801: Receive the electrode plate connection completion signal; The generation of the electrode connection completion signal includes: obtaining the impedance value of each electrode channel through an impedance detection circuit; generating a connection validity signal when the impedance value is within a preset threshold range; and verifying the electrode position matching based on the mapping relationship between the acupoint list and the electrode channel number. If the matching is successful, an electrode connection completion signal is output.

[0084] Specifically, the system can match the currently connected electrode channel number with the acupoint location specified in the treatment plan using a preset mapping table between acupoints and electrode channel numbers. For example, in an electro-moxibustion treatment plan, the acupoints "Zusanli" and "Hegu" are required to be stimulated, corresponding to electrode channels 1 and 3 respectively. If the system detects that channels 2 and 4 are currently connected, it determines that the electrode position is incorrect, the system will not continue the subsequent process, and will prompt the user to reattach the electrodes. This verification mechanism effectively avoids treatment errors caused by incorrect electrode placement, ensuring the accuracy and safety of the electrical pulse signal output.

[0085] Step S802: Analyze the channel allocation strategy in the electro-moxibustion therapy plan; the channel allocation strategy includes the waveform parameters, frequency parameters, and duration parameters of each target acupoint; The channel allocation strategy refers to the configuration information of the electrical pulse parameters specified for each target acupoint in the electro-moxibustion therapy plan, including but not limited to waveform type (such as square wave, triangular wave, sine wave), frequency parameters (such as 2Hz, 100Hz), and stimulation duration (such as 10 minutes, 30 minutes). These parameters are usually stored in the therapy plan file in a structured data format (such as JSON, XML). The system parses the file to extract the channel parameters corresponding to each acupoint and establishes a mapping relationship between acupoint, channel, and parameter. For example, for the "Zusanli" acupoint, its channel parameters are: waveform = square wave, frequency = 2Hz, duration = 20 minutes; while for the "Neiguan" acupoint, its channel parameters are: waveform = triangular wave, frequency = 10Hz, duration = 15 minutes. This parsing process is the basis for the subsequent generation of control instructions and is also the key to realizing personalized, multi-parameter electro-moxibustion treatment.

[0086] Step S803: Generate a channel control instruction set based on the channel allocation strategy; The channel control instruction set is a structured set of control commands used to guide the electrical pulse generator to independently control each channel according to set parameters. Each control instruction typically includes the following information: channel number, waveform type code, frequency value, stimulation duration, and output enable flag. For example, the system might generate the instruction: "CH1:SQUARE,2Hz,20min,ENABLE", indicating that a square wave with a frequency of 2Hz will be output to channel 1 for 20 minutes. This instruction set is usually dynamically generated by the system's central controller (such as an ARM Cortex-M series MCU) based on the parsing results and stored in memory for later transmission.

[0087] Step S804: The channel control instruction set is sent to the electrical pulse generator through the serial communication interface; wherein, the channel control instruction set is used to control the electrical pulse generator to output multi-channel electrical pulse signals to the target acupoint.

[0088] The electrical pulse generator is a system capable of emitting electrical pulse signals. Its core module is a function generator, which can provide multiple function signal outputs with adjustable waveforms, frequencies, and signal durations. In this embodiment, the electrical pulse generator is connected to an electrode pad adhered to an acupoint on the target object via an output cable. After the connection is completed, an automatic connection check is performed, and an electrode pad connection completion signal is output after confirmation.

[0089] Specifically, the caregiver manually attaches the electrode pads to the target acupoints according to the indication of the light spots, and connects the output cable of the multi-channel electrical pulse generator to the electrode pads. A line detection algorithm can be used to determine the correspondence between all connections and all acupoints requiring treatment (i.e., acupoints with light spot projections). After confirming the connections are correct, the electro-moxibustion control module reads the signal waveform type, signal frequency, and signal duration of the electrical pulse stimulation for each acupoint requiring treatment (i.e., acupoints with light spot projections) in the treatment plan; then, it sends commands through the communication interface to control the electrical pulse generator, outputting corresponding treatment signals to each acupoint, thus achieving the goal of treating and rehabilitating the patient.

[0090] It is understood that the intelligent electro-moxibustion therapy system provided in this application embodiment also includes functional modules such as a display module, a central processing unit, and a power supply module, and the various modules in the system are interconnected. The display module can be used to display various information such as acquired images, the identity information of the target object, acupoint recognition results, electro-moxibustion therapy plans, and projected images. The central processing unit is used to control the normal operation of each module. The power supply module is used to supply power to the system.

[0091] Furthermore, it should be noted that in the intelligent electro-moxibustion therapy control method based on acupoint recognition provided in this application, the network models involved in each step need to be trained before use. For specific training methods, please refer to the existing related model training process.

[0092] In the above embodiments, the reliability and positional accuracy of electrode connections are ensured. Through structured analysis of channel allocation strategies, personalized electrical pulse parameter configurations for different acupoints are achieved, which not only improves the precision and safety of electro-moxibustion therapy, but also realizes the intelligent upgrade of traditional electro-moxibustion treatment methods.

[0093] This application also discloses an intelligent electro-moxibustion physiotherapy control system based on acupoint recognition.

[0094] A smart electro-moxibustion therapy control system based on acupoint recognition, the control system comprising: The image acquisition module is used to acquire image information of the target object, including a first image for identification and a second image for acupoint identification. The identity recognition module is used to determine the identity information of the target object based on the first image; The electro-moxibustion therapy plan acquisition module is used to retrieve the electro-moxibustion therapy plan corresponding to the target object from a preset database based on the identity information. The acupoint recognition module is used to detect key points of acupoints based on the second image and output the acupoint recognition results. The positioning indicator module is used to guide the target subject into position based on the electro-moxibustion therapy plan; The projection control module is used to filter the coordinates of the target acupoints in the electro-moxibustion therapy plan from the acupoint recognition results and project light spots. The electro-moxibustion control module is used to control the output of multi-channel electrical pulse signals from the electrical pulse generator to the target acupoint.

[0095] An intelligent electro-moxibustion therapy control system based on acupoint recognition according to an embodiment of this application can implement any of the above-mentioned intelligent electro-moxibustion therapy control methods, and the specific working process of each module in the intelligent electro-moxibustion therapy control system can refer to the corresponding process in the above-mentioned method embodiments.

[0096] Reference Figure 9 This is a schematic diagram illustrating the application scenario of the intelligent electro-moxibustion therapy control system based on acupoint recognition provided in this application. The specific steps are as follows: Step 1: The image acquisition module is controlled to acquire full-body images of the patient lying on the standardized hospital bed, and the images are pre-processed by the image processing module. Step two: The image processing module sends the processed face image to the identity recognition module for automatic face detection and recognition to determine the patient's identity and retrieve the patient's electro-moxibustion therapy plan from the database; at the same time, the processed body image is sent to the acupoint recognition module for automatic acupoint recognition to obtain the acupoint recognition results of the current patient. Step 3: The positioning indicator module guides the target patient to position themselves according to the electro-moxibustion therapy plan and automatically determines the patient's position. Step 5: The projection control module selects the coordinates of the target acupoints according to the electro-moxibustion therapy plan and projects light spots onto the acupoints. Step six: Under the guidance of the projected light point, the caregiver operates the electrical pulse generator and manually connects it to the target acupoint. At the same time, the system automatically checks the connection and outputs a signal indicating that the electrode connection is complete after confirming that there are no errors.

[0097] Step 7: The electro-moxibustion control module controls the electric pulse generator to generate corresponding multi-channel electric pulse signals according to the electro-moxibustion therapy plan, so as to realize the electro-moxibustion therapy for the patient.

[0098] This application significantly improves the intelligence and automation level of electro-moxibustion therapy, reduces reliance on professional knowledge and manpower, promotes the development of the field of smart TCM, and provides strong support for the modernization and popularization of TCM therapy.

[0099] In the several embodiments provided in this application, it should be understood that the provided methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for example, the division of a certain module is merely a logical functional division, and in actual implementation there may be other division methods, such as multiple modules can be combined or integrated into another system, or some features can be ignored or not executed.

[0100] This application also discloses a computer-readable storage medium.

[0101] A computer-readable storage medium storing a computer program that can be loaded by a processor and executed as described above in any of the intelligent electro-moxibustion physiotherapy control methods based on acupoint recognition.

[0102] The computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in connection with an instruction execution system, apparatus, or device; the program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0103] It should be noted that the computer device and storage medium in the embodiments of this application are respectively electronic devices and storage media applying the above-described intelligent electro-moxibustion therapy control method. Therefore, all embodiments of the above-described intelligent electro-moxibustion therapy control method are applicable to the computer device and storage medium, and can achieve the same or similar beneficial effects. For the computer device / storage medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple; relevant details can be found in the descriptions of the method embodiments.

[0104] In this application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0105] Although this application has been described herein in conjunction with various embodiments, those skilled in the art, by reviewing the accompanying drawings, disclosure, and appended claims, will understand and implement other variations of the disclosed embodiments in carrying out the claimed application. In the claims, the word "comprising" does not exclude other components or steps, and "a" or "an" does not exclude a plurality. A single processor or other unit can implement several functions listed in the claims. While different dependent claims may recite certain measures, this does not mean that these measures cannot be combined to produce a good effect.

[0106] The above are all preferred embodiments of this application and are not intended to limit the scope of protection of this application. Any feature disclosed in this specification (including the abstract and drawings) may be replaced by other equivalent or similar features unless specifically stated otherwise. That is, unless specifically stated otherwise, each feature is only one example of a series of equivalent or similar features.

Claims

1. An intelligent electro-moxibustion physiotherapy control method based on acupoint recognition, characterized in that, The control method comprises: acquiring image information of a target object, the image information comprising a first image for identity recognition and a second image for acupoint recognition; determining identity information of the target object based on the first image; obtaining an electro-moxibustion therapy scheme corresponding to the target object from a preset database according to the identity information; performing acupoint key point detection based on the second image to output an acupoint recognition result; guiding the target object to a position based on the electro-moxibustion therapy scheme, screening target acupoint coordinates in the electro-moxibustion therapy scheme from the acupoint recognition result, and performing light point projection; controlling an electric pulse generator to output a multi-channel electric pulse signal to the target acupoint.

2. The intelligent electro-moxibustion therapy control method based on acupoint recognition according to claim 1, characterized in that, After the step of acquiring the image of the target object, the method further comprises: converting the image of the target object into a grayscale image; performing a normalization operation on the grayscale image to obtain a normalized image; adjusting the contrast of the normalized image based on a grayscale histogram equalization algorithm to obtain a preprocessed image, including a preprocessed first image and a second image.

3. The intelligent electro-moxibustion therapy control method based on acupoint recognition according to claim 1, characterized in that, The steps of determining the identity information of the target object based on the first image and obtaining the electro-moxibustion therapy scheme corresponding to the target object from the pre-stored database according to the identity information comprise: inputting the first image into a pre-trained face detection model for detection to obtain a face region image by cropping; inputting the face region image into a pre-trained face recognition model for recognition to output an identity ID as the identity information of the target object; searching the preset database based on the identity ID to match the electro-moxibustion therapy scheme corresponding to the identity ID.

4. The intelligent electro-moxibustion therapy control method based on acupoint recognition according to claim 1, characterized in that, The steps of performing acupoint key point detection based on the second image to output an acupoint recognition result comprise: inputting the second image into a ResNet50 backbone network to extract image features; performing key point coordinate regression on the image features through a nearest neighbor regression module to output a two-dimensional coordinate set of all acupoints to obtain the acupoint recognition result; wherein the nearest neighbor regression module is used to predict the main offset of each acupoint and the correlation offset of adjacent acupoints.

5. The intelligent electro-moxibustion therapy control method based on acupoint recognition according to any one of claims 1 to 4, characterized in that: The electro-moxibustion therapy scheme comprises an acupoint list, a channel allocation strategy, and a body position sequence; The steps of guiding the target object to a position based on the electro-moxibustion therapy scheme, screening target acupoint coordinates in the electro-moxibustion therapy scheme from the acupoint recognition result, and controlling the electric pulse generator to output a multi-channel electric pulse signal to the target acupoint comprise: generating body position adjustment prompts in sequence based on the body position sequence to guide the target object to adjust to target body positions in sequence; screening a target acupoint coordinate set from the acupoint recognition result according to the current body position state of the target object and the acupoint list; wherein the current body position state comprises supine, prone, left lateral recumbency, or right lateral recumbency; performing spatial mapping calculation according to the target acupoint coordinate set to generate a light point projection coordinate set; generating a black background image based on the target body position, and lighting corresponding position pixels on the black background image according to the light point projection coordinate set to form a projection image with light points; controlling a light point projection device to project the projection image to the body surface of the target object; Output a multi-channel electric pulse signal to the target acupoints based on the channel allocation strategy.

6. The intelligent electro-moxibustion therapy control method based on acupoint recognition according to claim 5, characterized in that, After the step of sequentially generating a body position adjustment prompt based on the body position sequence to guide the target object to sequentially adjust to a target body position in sequence, the method further comprises: Acquire a real-time body position image of the target object; Identify a body posture feature of the target object according to the real-time body position image, and determine a current body position state; Determine whether the current body position state conforms to a target body position in the body position sequence; if not, output a body position correction prompt.

7. The intelligent electro-moxibustion therapy control method based on acupoint recognition according to claim 5, characterized in that, The step of performing spatial mapping calculation based on the target acupoint coordinate set to generate a light point projection coordinate set comprises: Calculate a translation and rotation parameter based on a preset projection coordinate sample set and the target acupoint coordinate set, and the calculation formula is: (Δx,Δy,Δθ) T =(x d ,y d ,θ d ) T -(x D ,y D ,θ D ) T Where (Δx, Δy, Δθ) represent translation and rotation parameters, (x... d ,y d ,θ d (x) represents the preset projection coordinate sample. D ,y D ,θ D () represents the coordinates of the target acupoint, and T represents the matrix transpose symbol; Calculate a correction amount of the to-be-projected coordinate according to the translation and rotation parameter, and the calculation formula is: wherein (x i ,y i ,θ i ) represents the coordinate to be projected, represents the correction amount of the coordinate to be projected; Correct the target acupoint coordinate set based on the correction amount, and output a light point projection coordinate set.

8. The intelligent electro-moxibustion therapy control method based on acupoint recognition according to claim 5, characterized in that, Before the step of outputting a multi-channel electric pulse signal to the target acupoints based on the channel allocation strategy, the method further comprises: Analyze a channel allocation strategy in the electro-moxibustion therapy scheme; the channel allocation strategy comprises waveform parameters, frequency parameters and time length parameters of each target acupoint; Generate a channel control instruction set based on the channel allocation strategy; Send the channel control instruction set to an electric pulse generator through a serial communication interface; wherein the channel control instruction set is used to control the electric pulse generator to output a multi-channel electric pulse signal to the target acupoints.

9. An intelligent electro-moxibustion physiotherapy control system based on acupoint recognition, characterized in that, The control system comprises: An image acquisition module for acquiring image information of a target object, wherein the image information comprises a first image for identity recognition and a second image for acupoint recognition; An identity recognition module for determining identity information of a target object based on the first image; An electro-moxibustion therapy scheme acquisition module for acquiring an electro-moxibustion therapy scheme corresponding to the target object from a preset database according to the identity information; An acupoint recognition module for performing acupoint key point detection based on the second image, and outputting an acupoint recognition result; A positioning indication module for guiding a target object to position based on the electro-moxibustion therapy scheme; A projection control module for screening target acupoint coordinates in the electro-moxibustion therapy scheme from the acupoint recognition result and performing light point projection; An electro-moxibustion control module for controlling an electric pulse generator to output a multi-channel electric pulse signal to the target acupoints.

10. A computer-readable storage medium, characterized in that: A computer program is stored, which can be loaded and executed by a processor to perform the method of any one of claims 1 to 8.

Images