A human back acupoint calibration method and system based on real-time image optimization

Abstract

The application discloses a human back acupoint calibration method and system based on real-time image optimization, and relates to the fields of image processing and medical technology.The method comprises the following steps: collecting a human image and performing a pretreatment operation to obtain a pretreated image; performing a fusion operation on the pretreated image based on a multi-modal fusion image recognition algorithm to obtain a fused image; performing a correction operation on the human image based on multi-modal features in the fused image to obtain a corrected image; inputting the corrected image into a target detection model to obtain a human model image; performing a posture evaluation on the human model image to detect key point coordinate information on the human model image; correcting the key point coordinate information by using a key point correction model to obtain corrected key point coordinate information; and calculating acupoint coordinate information according to the corrected key point coordinate information.The application can calibrate the positions of human back acupoints in real time, efficiently and accurately.

Description

Technical Field

[0001] This invention relates to the fields of image processing and medical technology, and in particular to a method and system for calibrating acupoints on the human back based on real-time image optimization. Background Technology

[0002] Acupoint calibration is a popular task in traditional Chinese medicine and image processing, applicable to moxibustion, acupuncture, and massage. Traditional acupoint calibration methods rely on manual operation and professional experience, which are time-consuming and easily affected by subjective factors. In recent years, with the rapid development of computer vision and deep learning, acupoint calibration of the human back using image processing has become a new research direction and solution. However, ordinary two-dimensional images suffer from perspective distortion and lack of depth information due to distortion caused by viewing angle and shooting conditions, which may lead to inaccurate geometric information. Summary of the Invention

[0003] The purpose of this invention is to provide a method and system for calibrating acupoints on the human back based on real-time image optimization, which can calibrate the location of acupoints on the human back in real time, efficiently and accurately.

[0004] To achieve the above objectives, the present invention provides the following solution:

[0005] A method for calibrating acupoints on the human back based on real-time image optimization, comprising:

[0006] Human images are acquired and preprocessed to obtain preprocessed images;

[0007] A multimodal fusion image recognition algorithm is used to fuse the preprocessed image to obtain the fused image.

[0008] The human body image is corrected based on the multimodal features in the fused image to obtain the corrected image;

[0009] The corrected image is input into the target detection model to obtain the human body model image;

[0010] Pose assessment is performed on human body model images to detect the coordinate information of key points on the human body model images;

[0011] The key point coordinate information is corrected using a key point correction model to obtain the corrected key point coordinate information;

[0012] The acupoint coordinates are calculated based on the corrected key point coordinates.

[0013] Optionally, the step of acquiring human images and performing preprocessing operations to obtain preprocessed images specifically includes:

[0014] Video frames or human images are captured by a camera, and the image size is unified to 480×640 through data normalization to obtain a preprocessed image.

[0015] Optionally, the multimodal fusion-based image recognition algorithm performs a fusion operation on the preprocessed image to obtain a fused image, specifically including:

[0016] Collect multimodal data of the current environment; the multimodal data includes image data collected by infrared sensors, ambient light sensors and / or other cameras;

[0017] The wavelet transform method is used to fuse multimodal data with the preprocessed image to obtain the fused image.

[0018] Optionally, the step of correcting the human image based on the multimodal features in the fused image to obtain the corrected image specifically includes:

[0019] Extract multimodal features from the fused image;

[0020] Based on multimodal features, a 3D reconstruction algorithm is used to obtain the 3D reconstructed image;

[0021] The human body image is corrected based on the 3D reconstructed image to obtain the corrected image.

[0022] Optionally, the step of inputting the corrected image into the target detection model to obtain the human body model image specifically includes:

[0023] Construct an object detection model based on the YOLO object detection algorithm;

[0024] The corrected image is input into the target detection model, and the rectangular bounding box of the human body model is output.

[0025] The human body model is segmented from the rectangle using OpenCV's imagecrop function to obtain the human body model image.

[0026] Optionally, the step of performing pose evaluation on the human body model image and detecting the coordinate information of key points on the human body model image specifically includes:

[0027] The body method in OpenPose.Solution was used to detect the coordinates of 25 key points in the human body model image. These 25 key points include: nose, neck, left knee, right knee, left ankle, right ankle, left shoulder, right shoulder, left eye, right eye, left elbow, right elbow, left wrist, right wrist, left ear, right ear, left thumb, right thumb, left little finger, right little finger, left heel, right heel, left hip, right hip, and mid-hip.

[0028] The system determines whether the human body is facing away from the camera based on the coordinates of the left shoulder, right shoulder, left eye, and right eye.

[0029] If the human body is facing the camera directly, the detection ends.

[0030] If the human body is facing away from the camera, output the coordinate information of 25 key points.

[0031] Optionally, the step of correcting the key point coordinate information using a key point correction model to obtain corrected key point coordinate information specifically includes:

[0032] Use the Pose method in OpenPose.Solution to build a keypoint correction model based on CNN;

[0033] The left shoulder, right shoulder, left hip, right hip, midpoint of the shoulder, and midpoint of the femur are input into the key point correction model as key points for spine calculation to obtain the precise coordinate information of the key points for spine calculation.

[0034] Optionally, the step of calculating the acupoint coordinate information based on the corrected key point coordinate information specifically includes:

[0035] The position of the spine is calculated based on the precise coordinate information of the key points for spine calculation.

[0036] Based on the location of the spine, the corrected coordinates of key points, and the measurements determined by the height of the human body, the coordinates of acupoints on the human body image are calculated.

[0037] A system for calibrating acupoints on the human back based on real-time image optimization, comprising:

[0038] The image preprocessing module is used to acquire human images and perform preprocessing operations to obtain preprocessed images;

[0039] The image fusion module is used to perform a fusion operation on the preprocessed image based on a multimodal fusion image recognition algorithm to obtain the fused image.

[0040] The image correction module is used to perform correction operations on human images based on multimodal features in the fused image to obtain the corrected image;

[0041] The human body model detection module is used to input the corrected image into the target detection model to obtain the human body model image;

[0042] The human posture assessment module is used to assess the posture of human model images and detect the coordinate information of key points on the human model images.

[0043] The coordinate correction module is used to correct the coordinate information of key points using the key point correction model, so as to obtain the corrected coordinate information of key points.

[0044] The acupoint coordinate calculation module is used to calculate the acupoint coordinate information based on the corrected key point coordinate information.

[0045] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0046] This invention provides a method and system for calibrating acupoints on the human back based on real-time image optimization. For preprocessed human images or video frames, a multimodal fusion image recognition algorithm is used to optimize the human image, followed by correction using a 3D reconstructed image, resulting in a more accurate image. Subsequently, a target detection model constructed using the YOLO object detection algorithm is used to segment the human model, obtaining a human model image. The OpenPose algorithm is used to perform pose estimation and keypoint detection on the human model image, accurately detecting the pose and keypoints in real time, avoiding the manual calibration process in traditional methods. An artificial intelligence model (i.e., a keypoint correction model trained by machine learning) is used to correct the keypoints, obtaining accurate coordinate information and effectively improving detection accuracy. Finally, using anatomy, the location of back acupoints, and human structure, the accurate location of back acupoints is quickly determined through the relative relationship between specific locations and keypoints, and the calibration results are output or displayed. This invention combines computer vision and medical knowledge, possessing high reliability and practicality. Attached Figure Description

[0047] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1 A flowchart of the method for calibrating acupoints on the human back based on real-time image optimization provided by the present invention;

[0049] Figure 2 This is a schematic diagram showing the positions of 25 key points on a human body model image;

[0050] Figure 3 A schematic diagram illustrating the process of the method for calibrating acupoints on the human back based on real-time image optimization provided by the present invention;

[0051] Figure 4 This is a schematic diagram illustrating the pose estimation and acupoint calibration process of the human body model image according to the present invention. Detailed Implementation

[0052] 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.

[0053] The purpose of this invention is to provide a method and system for calibrating acupoints on the human back based on real-time image optimization, which can calibrate the location of acupoints on the human back in real time, efficiently and accurately.

[0054] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0055] See Figure 1 and Figure 3 This invention discloses a method for calibrating acupoints on the human back based on real-time image optimization, comprising:

[0056] Step 1: Acquire human body images and perform preprocessing operations to obtain preprocessed images.

[0057] This invention captures video frames containing human images using a camera or directly captures human images, and performs data normalization on the human images to unify the image size to 480×640, thus obtaining a preprocessed image.

[0058] Step 2: Perform a fusion operation on the preprocessed image based on the multimodal fusion image recognition algorithm to obtain the fused image.

[0059] Multimodal fusion image recognition algorithms combine human images from different perspectives and multimodal data collected by different sensors to provide more comprehensive information, improve the quality and detail of human images, enhance the detail preservation and clarity of human images, and effectively improve the results of object segmentation and 3D reconstruction.

[0060] To obtain more accurate human images and environmental information, facilitating subsequent human image correction and recognition, this invention collects multimodal data of the current environment. This multimodal data includes image data captured by infrared sensors, ambient light sensors, and / or other cameras. Subsequently, wavelet transform is used to fuse the multimodal data with the preprocessed image, resulting in a fused image. Specifically, wavelet transform is applied to each modality of data, and then the corresponding sub-bands are fused. Finally, inverse wavelet transform is used to obtain the fused image. Wavelet transform provides information at different frequencies and scales, enabling detailed fusion between different modalities.

[0061] Step 3: Perform a correction operation on the human body image based on the multimodal features in the fused image to obtain the corrected image.

[0062] OpenPose, an advanced open-source algorithm for human pose estimation, is based on deep learning and convolutional neural networks. It uses 3D reconstruction technology to reconstruct human and environmental images and correct distorted images to increase accuracy. It is also a method for accurate measurement, analysis, and simulation using 3D data and is often used in applications such as precise size detection, morphological analysis, and physical simulation.

[0063] Step 3 specifically includes:

[0064] Step 3.1: Extract multimodal features from the fused image. These multimodal features include feature points or feature descriptions, such as corner points, edges, and textures. In multi-view images, the multimodal features should maintain stability and consistency across different viewpoints.

[0065] Step 3.2: Based on multimodal features, obtain the 3D reconstructed image using a 3D reconstruction algorithm. Specifically, utilize matched multimodal features or disparity information to obtain the 3D reconstructed image using a 3D reconstruction algorithm.

[0066] Step 3.3: Correct the human body image based on the 3D reconstructed image to obtain the corrected image. Specifically, based on the shape of the 3D reconstructed image, optimize the details of the original human body image by adjusting the image using methods including rotation, translation, and scaling transformations to make it consistent with the actual image.

[0067] Step 4: Input the corrected image into the target detection model to obtain the human body model image.

[0068] YOLO is a popular real-time object detection algorithm that transforms the object detection task into a regression problem through a single forward pass, resulting in faster detection speed and better accuracy compared to traditional object detection algorithms.

[0069] Step 4 specifically includes:

[0070] Step 4.1: Construct an object detection model based on the YOLO object detection algorithm.

[0071] The weight file for the object detection model uses the official YOLOv7 weight file, and the parameter file includes: use_cuda=true, model depth depth_multiple=0.3, model width_multiple=0.5, class nc=1, and confidence threshold conf-thresh=0.35.

[0072] Step 4.2: Input the corrected image into the target detection model and output the bounding box of the human body model.

[0073] The core idea of ​​YOLO is to divide the input human image into a grid and predict the bounding box (i.e., the rectangular box of the human model) and category of the target in each grid. That is, for each bounding box, the category to which the target belongs is also predicted.

[0074] Step 4.3: Use OpenCV's imagecrop function to segment the human body model from the rectangular frame to obtain the human body model image. Then save the human body model locally.

[0075] Step 5: Perform pose assessment on the human body model image and detect the coordinate information of key points on the human body model image.

[0076] OpenPose is a computer image algorithm whose core idea is to achieve human pose estimation by combining deep learning and computer vision techniques. Based on convolutional neural networks (CNNs) and a multi-stage processing flow, it performs forward inference on the input image to output human pose information and the location of key points.

[0077] See Figure 4 Step 5 specifically includes:

[0078] Step 5.1: Using the `body` method in `OpenPose.Solution`, detect the coordinates of 25 key points in the human body model image. These 25 key points include: nose, neck, left knee, right knee, left ankle, right ankle, left shoulder, right shoulder, left eye, right eye, left elbow, right elbow, left wrist, right wrist, left ear, right ear, left thumb, right thumb, left little finger, right little finger, left heel, right heel, left hip, right hip, and mid-hip. The specific locations of these 25 key points on the human body image are as follows: Figure 2 As shown in Table 1, the numbers corresponding to each key point are shown in the table.

[0079] Table 125 Key Points Comparison Table

[0080]

[0081]

[0082] Step 5.2: Determine whether the human body is facing away from the camera based on the coordinates of the left shoulder, right shoulder, left eye, and right eye.

[0083] This invention primarily uses the coordinates of four key points—point 2 (right shoulder), 5 (left shoulder), 15 (right eye), and 16 (left eye)—to determine whether the human body in an image is facing away from the camera. Points 0 (nose), 17 (right ear), 18 (left ear), 4 (right wrist), and 7 (left wrist) assist in this determination. If the x-coordinate of point 2 is greater than that of point 5, and the x-coordinate of point 17 is greater than that of point 18, then the human body in the image is determined to be facing away from the camera; otherwise, the detection fails and ends.

[0084] Step 5.3: If the human body is facing the camera directly, the detection ends.

[0085] Step 5.4: If the human body is facing away from the camera, output the coordinate information of 25 key points.

[0086] Step 6: Use the key point correction model to correct the key point coordinate information to obtain the corrected key point coordinate information.

[0087] Step 6 specifically includes:

[0088] Step 6.1: Use the Pose method in OpenPose.Solution to build a keypoint correction model based on CNN.

[0089] The keypoint correction model of this invention uses a CNN model. The training set consists of common human back images. First, OpenPose is used to label the images, and then the coordinates of the right shoulder, left shoulder, right hip, left hip, midpoint of the right shoulder and left shoulder, and midpoint of the right hip and left hip are labeled as input for training the keypoint correction model, ensuring an accuracy of over 85%.

[0090] Step 6.2: Input the left shoulder, right shoulder, left hip, right hip, midpoint of the shoulder, and midpoint of the femur as key points for spine calculation into the key point correction model to obtain the precise coordinate information of the key points for spine calculation.

[0091] After determining the human body posture in the human body image, that is, when the human body posture in the human body image is facing away from the camera, after outputting the key point coordinate information, the machine learning (key point correction model) is used to correct the key point coordinate information. The model inputs the coordinate information of key points 2 (right shoulder), 5 (left shoulder), 9 (right hip), 12 (left hip), the midpoint between 2 (right shoulder) and 5 (left shoulder), and the midpoint between 9 (right hip) and 12 (left hip), and outputs the accurate coordinates of the right shoulder, left shoulder, right hip, left hip, scapula and femur, which are the accurate coordinate information of the key points for spine calculation. Subsequently, the accurate coordinate information of the key points for spine calculation is substituted into the straight line formula \((X - X1) / (X2 - X1)=(Y - Y1) / (Y2 - Y1)\) to obtain the spine function, where X is the abscissa of the spine position, Y is the ordinate of the spine position, X1 is the abscissa of the midpoint between the right shoulder and the left shoulder, Y1 is the ordinate of the midpoint between the right shoulder and the left shoulder, X2 is the abscissa of the midpoint between the right hip and the left hip, and Y2 is the ordinate of the midpoint between the right hip and the left hip.

[0092] Step 7: Calculate the acupoint coordinate information according to the corrected key point coordinate information.

[0093] The specific steps of step 7 include:

[0094] Step 7.1: Calculate the spine position according to the accurate coordinate information of the key points for spine calculation.

[0095] Substitute the accurate coordinate information of the key points for spine calculation into the spine function, and the spine position can be calculated.

[0096] Step 7.2: Calculate the coordinate information of each acupoint on the human body image according to the spine position, key point coordinate information and the measurement in inches determined by the human body height.

[0097] Determine the measurement in inches of the human body model image according to GB / T12346-2021 "Names and Locations of Acupoints". Set the human body height to 75 equal parts according to the bone measurement in inches in the national standard 3.3.1, and measure the remaining position relationships and widths according to this ratio. Specific method: Set the human body height to 75 equal parts, and then fold the human body model image into a certain number of equal parts according to a certain section length and width. One part is one inch, which is the main basis for acupoint location. Approximately regard the midpoint between the right shoulder and the left shoulder as the Taodao acupoint, and approximately regard the midpoint between the right hip and the left hip as the Yaoshu acupoint.

[0098] After adjustments based on actual human proportions, and using the established dimensions, spine, and key point coordinates, the coordinates of various acupoints on the back are derived according to GB / T12346-2021 "Names and Locations of Acupoints" through translation transformations and coordinate addition and subtraction. For example, according to the national standard, the spine is located between the Dazhui (GV14) and Yaoshu (GV26) acupoints, and the Taodao (GV14) acupoint is one cun below the Dazhui acupoint. Therefore, the coordinates of the Dazhui acupoint are obtained by subtracting the established dimensions from the coordinates of the Taodao acupoint. Additionally, the Feishu (BL13) acupoint is located 3 cun below the Taodao acupoint and 1.5 cun lateral to it on the back; the Fengmen (BL12) acupoint is located on the spine, 1.5 cun lateral to the posterior midline below the spinous process of the second thoracic vertebra, and so on. Finally, the coordinates of various acupoints on the human body image are calculated and output or displayed.

[0099] In summary, the method for calibrating acupoints on the human back based on real-time image optimization provided by this invention can accurately detect key points in human images or video frames, including key points in the head, arms, legs, and back. Based on traditional Chinese medicine theory and anatomical knowledge, the method infers the accurate location of acupoints on the back by analyzing the spatial distribution and positional relationship of key points. It has the characteristics of high speed, high accuracy, and wide applicability.

[0100] Based on the aforementioned method for calibrating acupoints on the human back using real-time image optimization, this invention also provides a system for calibrating acupoints on the human back using real-time image optimization, comprising:

[0101] The image preprocessing module is used to acquire human images and perform preprocessing operations to obtain preprocessed images.

[0102] The image fusion module is used to perform fusion operations on preprocessed images based on a multimodal fusion image recognition algorithm to obtain a fused image.

[0103] The image correction module is used to correct human images based on multimodal features in the fused image to obtain the corrected image.

[0104] The human body model detection module is used to input the corrected image into the target detection model to obtain the human body model image.

[0105] The human posture assessment module is used to assess the posture of human model images and detect the coordinate information of key points on the human model images.

[0106] The coordinate correction module is used to correct the coordinate information of key points using the key point correction model, so as to obtain the corrected coordinate information of key points.

[0107] The acupoint coordinate calculation module is used to calculate the acupoint coordinate information based on the corrected key point coordinate information.

[0108] Compared with existing methods for calibrating acupoints, the method and system for calibrating human back acupoints based on real-time image optimization provided by this invention have at least the following advantages:

[0109] 1) The method and system for calibrating acupoints on the back of the human body based on real-time image optimization provided by this invention uses the YOLO target detection algorithm to construct a target detection model to segment the human body model. The classification only involves the human body, which makes the processing process simpler and significantly improves the processing efficiency.

[0110] 2) The method and system for calibrating acupoints on the back of the human body based on real-time image optimization provided by this invention and the multimodal fusion image recognition algorithm used are effective optimization methods to improve the quality of human body images, enhance details, or achieve specific goals, and can provide more comprehensive and accurate image optimization results.

[0111] 3) The method and system for calibrating acupoints on the back of the human body based on real-time image optimization provided by this invention uses three-dimensional reconstruction to correct human body images. By obtaining an accurate three-dimensional human body model, it helps in human posture analysis and can provide more accurate and realistic human body image data.

[0112] 4) The method and system for calibrating acupoints on the human back based on real-time image optimization provided by this invention uses the coordinates of key points detected by OpenPose as reference points, which is faster, more accurate, and more in line with human anatomy.

[0113] 5) The method and system for calibrating acupoints on the human back based on real-time image optimization provided by this invention uses national standards to determine the measurement. By determining different measurement for each human body, this invention has wider applicability, greater versatility, and higher accuracy.

[0114] 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 systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.

[0115] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for calibrating acupoints on the human back based on real-time image optimization, characterized in that, include: Human images are acquired and preprocessed to obtain preprocessed images; A multimodal fusion image recognition algorithm is used to fuse the preprocessed image to obtain the fused image. Specifically, this includes: collecting multimodal data of the current environment; the multimodal data includes image data collected by infrared sensors, ambient light sensors, and / or other cameras; and using wavelet transform to fuse the multimodal data with the preprocessed image to obtain a fused image. The human body image is corrected based on the multimodal features in the fused image to obtain a corrected image. Specifically, this includes: extracting multimodal features from the fused image; these multimodal features include feature points or feature descriptions, such as corner points, edges, and textures; using a 3D reconstruction algorithm based on the multimodal features to obtain a 3D reconstructed image; specifically, using matched multimodal features or disparity information, a 3D reconstruction algorithm is used to obtain a 3D reconstructed image; the human body image is corrected based on the 3D reconstructed image to obtain a corrected image; specifically, based on the shape of the 3D reconstructed image, the details of the original human body image are optimized by adjusting the image using methods including rotation, translation, and scaling transformations. The corrected image is input into the target detection model to obtain the human body model image; Pose assessment is performed on human body model images to detect the coordinate information of key points on the human body model images; The key point coordinate information is corrected using a key point correction model to obtain the corrected key point coordinate information; The acupoint coordinates are calculated based on the corrected key point coordinates.

2. The method for calibrating acupoints on the human back based on real-time image optimization according to claim 1, characterized in that, The process of acquiring human images and performing preprocessing operations to obtain preprocessed images specifically includes: Video frames or human images are captured by a camera, and the image size is unified to 480×640 through data normalization to obtain a preprocessed image.

3. The method for calibrating acupoints on the human back based on real-time image optimization according to claim 1, characterized in that, The step of inputting the corrected image into the target detection model to obtain the human body model image specifically includes: Construct an object detection model based on the YOLO object detection algorithm; The corrected image is input into the target detection model, and the rectangular bounding box of the human body model is output. The human body model is segmented from the rectangle using OpenCV's imagecrop function to obtain an image of the human body model.

4. The method for calibrating acupoints on the human back based on real-time image optimization according to claim 1, characterized in that, The process of performing pose evaluation on the human body model image and detecting the coordinate information of key points on the human body model image specifically includes: The body method in OpenPose.Solution was used to detect the coordinates of 25 key points in the human body model image. These 25 key points include: nose, neck, left knee, right knee, left ankle, right ankle, left shoulder, right shoulder, left eye, right eye, left elbow, right elbow, left wrist, right wrist, left ear, right ear, left thumb, right thumb, left little finger, right little finger, left heel, right heel, left hip, right hip, and mid-hip. The system determines whether the human body is facing away from the camera based on the coordinates of the left shoulder, right shoulder, left eye, and right eye. If the human body is facing the camera directly, the detection ends. If the human body is facing away from the camera, output the coordinate information of 25 key points.

5. The method for calibrating acupoints on the human back based on real-time image optimization according to claim 4, characterized in that, The process of correcting the keypoint coordinate information using a keypoint correction model to obtain the corrected keypoint coordinate information specifically includes: Use the Pose method in OpenPose.Solution to build a keypoint correction model based on CNN; The left shoulder, right shoulder, left hip, right hip, midpoint of the shoulder, and midpoint of the femur are input into the key point correction model as key points for spine calculation to obtain the precise coordinate information of the key points for spine calculation.

6. The method for calibrating acupoints on the human back based on real-time image optimization according to claim 5, characterized in that, The calculation of acupoint coordinate information based on the corrected key point coordinate information specifically includes: The position of the spine is calculated based on the precise coordinate information of the key points for spine calculation. Based on the location of the spine, the corrected coordinates of key points, and the measurements determined by the height of the human body, the coordinates of acupoints on the human body image are calculated.

7. A system for calibrating acupoints on the human back based on real-time image optimization, characterized in that, include: The image preprocessing module is used to acquire human images and perform preprocessing operations to obtain preprocessed images; The image fusion module is used to perform a fusion operation on the preprocessed image based on a multimodal fusion image recognition algorithm to obtain the fused image. Specifically, this includes: collecting multimodal data of the current environment; the multimodal data includes image data collected by infrared sensors, ambient light sensors, and / or other cameras; and using wavelet transform to fuse the multimodal data with the preprocessed image to obtain a fused image. The image correction module is used to correct a human image based on multimodal features in the fused image to obtain a corrected image. Specifically, it includes: extracting multimodal features from the fused image; the multimodal features include feature points or feature descriptions, such as corner points, edges, and textures; using a 3D reconstruction algorithm based on the multimodal features to obtain a 3D reconstructed image; specifically, using matched multimodal features or disparity information, a 3D reconstruction algorithm is used to obtain a 3D reconstructed image; correcting the human image based on the 3D reconstructed image to obtain a corrected image; specifically, optimizing the details of the original human image based on the shape of the 3D reconstructed image, adjusting the image using methods including rotation, translation, and scaling transformations. The human body model detection module is used to input the corrected image into the target detection model to obtain the human body model image; The human posture assessment module is used to assess the posture of human model images and detect the coordinate information of key points on the human model images. The coordinate correction module is used to correct the coordinate information of key points using the key point correction model, so as to obtain the corrected coordinate information of key points. The acupoint coordinate calculation module is used to calculate the acupoint coordinate information based on the corrected key point coordinate information.

Images