Method and device for enhancing precision of tail joint of unmarked motion capture system

Abstract

The invention discloses an end joint precision enhancement method and device of an unmarked motion capture system, and relates to the technical field of human motion capture, and the method comprises the steps that a main system collects original motion capture data to obtain a joint initial pose of the whole body of a detected object; an enhancement module is installed at the tail end joint of the measured object, and the enhancement module at least comprises an optical mark point and an inertial sensor; performing spatial position identification on the tail end joint of the detected object to obtain first tail end joint position data, and synchronously acquiring second tail end joint position data of the enhancement module; performing fusion filtering processing on the first tail end joint position data and the second tail end joint position data to obtain high-precision position data of the tail end joint; and the enhanced tail end joint motion data of the enhancement module is collected, and the target tail end joint motion data is output by combining the enhanced tail end joint motion data and the high-precision position data. The method has the effect of improving the accuracy and robustness of the motion capture system.

Description

Technical Field

[0001] This application relates to the technical field of human motion capture, and in particular to a method and apparatus for enhancing the end-joint accuracy of a markerless motion capture system. Background Technology

[0002] Existing motion capture systems are broadly categorized into labeled and labelless systems. Labeled systems achieve high accuracy by attaching markers to key body parts and using infrared cameras for tracking, but require wearable markers, resulting in higher costs. Labelless motion capture systems rely on cameras and image algorithms to identify key points in the human skeleton, offering advantages such as ease of use and freedom from marker constraints. However, due to the small area of ​​end joints (such as the wrist and ankle) in images, their susceptibility to occlusion, and the high speed of movement, existing labelless systems often struggle to accurately estimate their positions, leading to decreased motion capture accuracy. For example, during rapid hand waving or foot movements, detection errors for end joints such as fingers and toes are significant. Current technologies lack an effective solution that simultaneously balances convenience and end joint capture accuracy. Therefore, a technical solution is urgently needed to improve the positioning accuracy of end joints in labelless motion capture systems. Summary of the Invention

[0003] The purpose of this invention is to provide a method and apparatus for enhancing the end-joint accuracy of a markerless motion capture system, so as to solve the problems mentioned in the background art.

[0004] In a first aspect, this application provides a method for enhancing the end-joint accuracy of a markerless motion capture system, the method comprising: The main system collects the raw motion capture data of the tested object, performs data recognition and filtering on the raw motion capture data, and obtains the preliminary joint pose of the tested object's whole body. An enhancement module is installed at one or more end joints of the object under test, and the enhancement module includes at least one optical marker and an inertial sensor; Based on the preliminary joint pose, the spatial position of the end joint of the tested object is identified to obtain the first end joint position data, and the second end joint position data of the enhancement module is collected simultaneously. The first end joint position data and the second end joint position data are fused and filtered to obtain high-precision position data of the end joint. The enhanced end joint motion data of the enhancement module is collected, and the target end joint motion data is output by combining the enhanced end joint motion data and the high-precision position data.

[0005] Preferably, the step of performing data identification and filtering on the raw motion capture data to obtain the preliminary joint poses of the whole body of the test subject specifically includes: Based on the original motion capture data, joint recognition is performed on the original motion capture data to obtain the joint locations of the tested object; The original motion capture data is filtered according to the joint location to obtain the initial joint motion data corresponding to each joint location. Construct a spatial coordinate system, and in the spatial coordinate system, draw the joint motion trajectory and joint movement changes according to the initial joint motion data; The joint posture is obtained based on the changes in joint movement, the joint spatial position is obtained based on the joint motion trajectory, and the preliminary joint pose is obtained by combining the joint posture and the joint spatial position.

[0006] Preferably, the step of performing spatial position recognition on the distal joint of the tested object based on the preliminary joint pose to obtain the first distal joint position data specifically includes: Based on the preliminary joint pose, the joint type corresponding to each preliminary joint pose is identified, and the joint type is filtered to obtain the preliminary pose of the target joint corresponding to the end joint. Based on the preliminary pose of the target joint, extract the joint spatial position in the preliminary pose of each target joint; Extract the target joint type of the distal joint, generate a joint label based on the target joint type, and add the joint label to the joint spatial position to obtain the first distal joint position data.

[0007] Preferably, the step of collecting the position data of the second end joint of the enhancement module specifically includes: Based on the optical markers, the camera captures images of the optical markers to obtain spatial location data of the optical markers; Based on the inertial sensor, the inertial sensor records the acceleration and angular velocity of the end joint; Based on the acceleration and the angular velocity, the velocity-angle change data of the end joint during the movement are obtained; The spatial position data of the marker points and the velocity-angle change data are fused according to a preset first weight ratio to obtain the second end joint position data of the end joint of the tested object, wherein the weight of the spatial position data of the marker points is greater than the weight of the velocity-angle change data.

[0008] Preferably, before the step of acquiring the position data of the second end joint of the enhancement module, the method further includes: Extract the data acquisition frequency of the main system during the process of collecting the raw motion capture data; The enhancement module records data synchronously with the main system based on the data acquisition frequency.

[0009] Preferably, the step of fusing and filtering the first end joint position data and the second end joint position data to obtain high-precision position data of the end joint specifically includes: Based on the first end joint position data, extract the first time-series data column of the first end joint position data; Based on the second end joint position data, extract the second time-series data column of the second end joint position data; The first distal joint position data is evaluated for clarity value to obtain the clarity value index of the first distal joint position data; Based on the clarity index, the first end joint position data and the second end joint position data are dynamically weighted to obtain a weight ratio. The first time-series data column and the second time-series data column are aligned, and the first end joint position data and the second end joint position data are fused according to the weight ratio to obtain the high-precision position data of the end joint.

[0010] Preferably, after the step of fusing the first end joint position data and the second end joint position data, the method further includes: Extract the full-body posture of the subject from the original motion capture data, and record the first end joint position data and the fused data after fusing the second end joint position data; The fused data is matched with the whole-body posture to obtain the matching value between the whole-body posture of the tested object and the fused data; The local precision value of the fused data is evaluated based on the matching value, and it is determined whether the local precision value is higher than a preset precision threshold. If it is determined that the local precision value is lower than the precision threshold, then a precision error correction parameter is generated based on the matching value, and the precision of the fused data is corrected according to the precision error correction parameter.

[0011] Preferably, the step of collecting the enhanced end-joint motion data of the enhancement module, combining the enhanced end-joint motion data with the high-precision position data, and outputting the target end-joint motion data specifically includes: The spatial position data of the marker points and the velocity-angle change data are fused according to a preset second weight ratio to obtain the enhanced end joint motion data of the tested object, wherein the weight of the spatial position data of the marker points is less than the weight of the velocity-angle change data; Based on the high-precision position data, the trajectory change data of the end joint of the tested object in space is obtained; Based on the enhanced joint motion data, the velocity change data of the end joint of the tested object in space is obtained; By combining the trajectory change data and the velocity change data, the motion data of the target end joint of the tested object is output.

[0012] Secondly, this application provides a device for enhancing the end-joint precision of a markerless motion capture system, the device comprising: Main system module: used to acquire video images and extract key points of the human skeleton to obtain the raw motion capture data of the tested object. The main system module includes at least two cameras that acquire data simultaneously. Enhancement module: used to be installed at the end joint of the object under test, and to collect second end joint position data and enhanced end joint position data, wherein the end joint includes the wrist, fingertips, ankle and toe tips, and the enhancement module includes at least one optical marker and an inertial sensor; Fusion module: Used to fuse the data output by the main system module and the enhancement module, and output the fused target end joint motion data.

[0013] In summary, this application includes at least one of the following beneficial technical effects: By incorporating an end-joint enhancement module into a labelless motion capture system and utilizing a data fusion algorithm to fuse the acquired data, the estimation accuracy of end-joint positions is significantly improved. Compared to traditional solutions that rely solely on camera recognition, this invention effectively reduces capture errors in end-joint areas such as the wrist and ankle, and can handle rapid movements and occlusion. Because the enhancement module is small, convenient, and can wirelessly transmit data, the system maintains the simplicity of labelless capture while providing higher accuracy and more reliable motion capture results in scenarios such as virtual reality, sports training, and human-computer interaction, thus improving the accuracy and robustness of the motion capture system. Attached Figure Description

[0014] Figure 1 This is a flowchart illustrating the steps of a method for enhancing the end-joint accuracy of a markerless motion capture system according to an embodiment of this application. Figure 2 This is a block diagram of a markerless motion capture system end-joint precision enhancement device provided in an embodiment of this application.

[0015] Explanation of reference numerals in the attached diagram: 1. Main system module; 2. Enhancement module; 3. Integration module. Detailed Implementation

[0016] The following is in conjunction with the appendix Figures 1-2 This application will be described in further detail, but the embodiments of the present invention are not limited thereto.

[0017] This application discloses a method and apparatus for enhancing the end-joint accuracy of a markerless motion capture system.

[0018] In this embodiment, a method for enhancing the end-joint accuracy of a markerless motion capture system includes: S100: The main system collects the raw motion capture data of the tested object, performs data recognition and filtering on the raw motion capture data, and obtains the preliminary joint pose of the tested object's whole body. S200: An enhancement module is installed at one or more end joints of the object under test. The enhancement module includes at least one optical marker and an inertial sensor. S300: Based on the preliminary joint pose, spatial position recognition is performed on the end joints of the tested object to obtain the first end joint position data, and the second end joint position data of the enhancement module is acquired simultaneously. S400: The first and second end joint position data are fused and filtered to obtain high-precision end joint position data; S500: Acquires enhanced end-joint motion data from the enhancement module, combines the enhanced end-joint motion data with high-precision position data, and outputs target end-joint motion data.

[0019] It should be noted that the above process is only the basic steps of this embodiment. In the specific implementation process, some steps may be added, reduced or modified appropriately without affecting the overall implementation effect.

[0020] The steps for identifying and filtering raw motion capture data to obtain the preliminary joint poses of the subject's entire body are as follows: Based on the raw motion capture data, joint recognition is performed on the raw motion capture data to obtain the joint locations of the tested object; The raw motion capture data is filtered based on the joint location to obtain the initial joint motion data corresponding to each joint location. Construct a spatial coordinate system, and in the spatial coordinate system, draw the joint motion trajectory and joint movement changes according to the initial joint motion data; The joint posture is obtained based on the changes in joint movement, the joint spatial position is obtained based on the joint motion trajectory, and the initial joint pose is obtained by combining the joint posture and the joint spatial position.

[0021] In practice, the main system uses two Intel RealSense D455 cameras to capture raw motion data of athletes on the treadmill at a frequency of 60Hz. When the athlete starts running, the system records raw data frames, each containing a 1920×1080 resolution RGB image and depth information. First, the system performs joint recognition on the raw data: using the OpenPose algorithm, 18 key points on the athlete's body are detected, including joints such as the shoulder (joint ID-S1), elbow (E1), wrist (W1), knee (K1), and ankle (A1). Then, the data is filtered according to the joint: for example, only the initial motion data of the wrist joint (W1) is extracted, including positional changes (X-axis displacement ±10mm, Y-axis displacement ±8mm). Next, a spatial coordinate system is constructed (origin set at the center of the treadmill), and the joint motion trajectory is plotted in the coordinate system: the athlete's wrist joint motion trajectory is displayed as an arc path with a total length of 1.2 meters over 5 seconds, and the movement changes include the wrist angle increasing from 0° to 30°. Based on the movement changes, the system calculates the joint posture: the wrist rotation angle error is ±2°, and the spatial position moves from the initial point (X=1.0m, Y=0.5m, Z=0.0m) to the end point (X=1.2m, Y=0.8m, Z=0.1m). Finally, combining the posture and position, the preliminary joint pose is obtained: the wrist pose data includes the timestamp 2023-09-15T10:00:00.000, coordinates (X=1.1m, Y=0.6m, Z=0.05m), angle 30°, and confidence level 85%.

[0022] The steps for identifying the spatial position of the distal joints of the tested object based on the preliminary joint pose and obtaining the first distal joint position data are as follows: Based on the initial joint pose, the joint type corresponding to each initial joint pose is identified, and the target joint initial pose corresponding to the end joint is obtained by filtering according to the joint type. Based on the initial pose of the target joint, extract the joint spatial position in the initial pose of each target joint; Extract the target joint type of the end joint, generate joint labels based on the target joint type, add the joint labels to the joint spatial location, and obtain the first end joint position data.

[0023] In application, based on preliminary joint pose data, the system identifies the joint type corresponding to each pose: for example, it detects joint ID-W1 as the wrist (end-stage joint), ID-A1 as the ankle (end-stage joint), and ID-S1 as the shoulder (non-end-stage joint). Then, it filters based on joint type: only the target preliminary poses of end-stage joints such as the wrist (W1) and ankle (A1) are selected. For example, when an athlete is running, the wrist pose data includes spatial position (X=1.1m, Y=0.6m, Z=0.05m) and posture angle of 30°, while the ankle pose data includes position (X=0.9m, Y=0.3m, Z=0.0m) and posture angle of 15°. Next, the joint spatial positions in these target preliminary poses are extracted: the wrist position data is (X=1.1m, Y=0.6m, Z=0.05m), and the ankle position data is (X=0.9m, Y=0.3m, Z=0.0m). Then, the target joint type of the distal joint is extracted: the wrist is labeled "Wrist", and the ankle is labeled "Ankle". Finally, the labels are added to the spatial location: the first distal joint position data is obtained, such as the timestamp 2023-09-15T10:00:00.005, the wrist data is {label: Wrist, position: (1.1, 0.6, 0.05)}, and the ankle data is {label: Ankle, position: (0.9, 0.3, 0.0)}, with an error range of ±3mm.

[0024] The specific steps for acquiring the position data of the second end joint of the enhancement module are as follows: Based on optical markers, the camera captures images of the optical markers to obtain spatial location data of the optical markers; Based on inertial sensors, the inertial sensors record the acceleration and angular velocity of the end joint; Based on acceleration and angular velocity, the velocity-angle change data of the end joint during the motion process are obtained; The spatial position data of the marker points and the velocity-angle change data are fused according to a preset first weight ratio to obtain the second end joint position data of the end joint of the tested object. The weight of the spatial position data of the marker points is greater than the weight of the velocity-angle change data.

[0025] In application, the enhancement modules are installed on the athlete's wrist and ankle: an optical marker (5mm in diameter, 90% reflectivity) is located on the wrist, and an inertial sensor (100Hz sampling rate) is located on the ankle. First, a camera (such as the main system's camera) captures an image of the optical marker: as the athlete runs, the camera captures the image of the marker, generating spatial location data; for example, the wrist marker's position at timestamp 2023-09-15T10:00:00.010 is (X=1.12m, Y=0.62m, Z=0.06m), with an error of ±2mm. Then, the inertial sensor records the acceleration and angular velocity of the distal joint: the wrist sensor measures an acceleration of 3.5m / s² and an angular velocity of 15rad / s, while the ankle sensor measures an acceleration of 4.0m / s² and an angular velocity of 10rad / s. Next, based on acceleration and angular velocity, the system calculates velocity-angle change data: the wrist velocity increases from 0 m / s to 0.35 m / s and the angle changes from 0° to 5° within 0.1 seconds; the ankle velocity increases from 0 m / s to 0.4 m / s and the angle changes from 0° to 3°. Finally, the spatial position data of the marker points and the velocity-angle change data are fused according to a first weight ratio: the weight ratio is 70% for the marker point data and 30% for the velocity-angle data. For example, after fusion, the second distal joint position data is obtained, with a timestamp of 2023-09-15T10:00:00.015, the wrist position is (X=1.115m, Y=0.615m, Z=0.055m), and the ankle position is (X=0.905m, Y=0.305m, Z=0.002m), with an accuracy improved to ±1.5mm.

[0026] Before the step of acquiring the position data of the second end joint of the enhancement module, the following steps are also included: Extract the data acquisition frequency of the main system during the process of acquiring raw motion capture data; The enhancement module records data synchronously with the main system based on the data acquisition frequency.

[0027] In application, before acquiring the second-end joint position data, the system extracts the data acquisition frequency of the main system: the main system camera acquires raw motion capture data at a frequency of 60Hz, with each frame spaced 16.7ms apart. Then, the enhancement module synchronously records data based on this frequency: for example, the sampling rate of the inertial sensor and optical markers is adjusted to 60Hz, synchronized with the main system's time. When the athlete begins accelerating at 10:00:00.020, the main system records one frame of data, with a timestamp accurate to milliseconds. Simultaneously, the enhancement module triggers: the exposure time of the optical marker camera is set to 16ms, and the inertial sensor data acquisition points are aligned to the same timestamp. In this way, all data is consistent on the timeline, with errors controlled within ±1ms, ensuring error-free subsequent fusion.

[0028] The steps for fusing and filtering the first and second end-joint position data to obtain high-precision end-joint position data are as follows: Based on the first distal joint position data, the first time-series data column of the first distal joint position data is extracted; Based on the second distal joint position data, extract the second time-series data column of the second distal joint position data; The clarity value of the first distal joint position data is evaluated to obtain the clarity value index of the first distal joint position data. The first and second distal joint position data are dynamically weighted according to the clarity index to obtain the weight ratio. The first and second time-series data columns are aligned, and the first and second end joint position data are fused according to the weight ratio to obtain high-precision end joint position data.

[0029] In application, the first time series data column is extracted based on the first distal joint position data (from the main system). For example, the sequence of wrist position at 5 time points (10:00:00.025 to 10:00:00.045) is [(1.10, 0.60, 0.05), (1.11, 0.61, 0.06), (1.12, 0.62, 0.07), (1.13, 0.63, 0.08), (1.14, 0.64, 0.09)]. Then, based on the second distal joint position data (from the enhancement module), a second time-series data column is extracted: the wrist position at the same time point is [(1.115, 0.615, 0.055), (1.125, 0.625, 0.065), (1.135, 0.635, 0.075), (1.145, 0.645, 0.085), (1.155, 0.655, 0.095)]. Next, the first distal joint position data is evaluated for data sharpness: a sharpness index is calculated. For example, if the sharpness (based on image quality) of the wrist data at the third time point (10:00:00.035) is 85% (threshold 80%), the index is 0.85. Then, weights are dynamically allocated according to the sharpness index: when the sharpness is ≥80%, 70% is given to the first data and 30% to the second data; otherwise, the weights are 50%:50%. Here, the weighting ratio is 70:30. Finally, the time-series data columns are aligned (using linear interpolation) and fused according to the weighting ratio to obtain high-precision location data. For example, for a timestamp of 10:00:00.035, the wrist position is (1.130.7 + 1.1350.3, 0.630.7 +0.6350.3, 0.080.7 +0.0750.3) = (1.1315, 0.6315, 0.0785), with an error of ±0.8mm.

[0030] After fusing the first and second distal joint position data, the method further includes: Extract the full-body pose of the subject from the original motion capture data, and record the first end joint position data and the fused data after fusing the second end joint position data; The fused data is matched with the full-body pose to obtain the matching value between the full-body pose of the tested object and the fused data; The local precision value of the fused data is evaluated based on the matching value, and it is determined whether the local precision value is higher than the preset precision threshold. If it is determined that the local precision value is lower than the precision threshold, a precision error correction parameter is generated based on the matching value, and the precision of the fused data is corrected according to the precision error correction parameter.

[0031] In application, after fusing the first and second distal joint position data, the system extracts the athlete's full-body posture from the original motion capture data. For example, at timestamp 2023-09-15T10:00:00.050, the full-body posture includes all joint positions, such as the shoulder (1.0, 0.7, 0.0) and hip (0.95, 0.4, 0.0). Then, the fused data is recorded: the wrist fused position is (1.1315, 0.6315, 0.0785). Next, the fused data is matched with the full-body posture: the matching value is calculated, such as the distance between the wrist and shoulder should be 0.3m (theoretical value), the actual distance is 0.3m, and the matching value is 95% (threshold 90%). Then, the local accuracy value is evaluated based on the matching value: the wrist accuracy value is 95%, which is higher than the threshold 90%. When the accuracy value is below a threshold (e.g., assuming a matching value of 85% at another time point), the system generates accuracy error correction parameters: for example, offset correction (ΔX=+0.005m, ΔY=+0.003m). Finally, the fused data is corrected according to the parameters: for example, the corrected position is (1.1315+ 0.005, 0.6315 +0.003, 0.0785) = (1.1365, 0.6345, 0.0785), ensuring improved accuracy.

[0032] The steps for acquiring enhanced end-joint motion data from the enhancement module, combining the enhanced end-joint motion data with high-precision position data, and outputting target end-joint motion data are as follows: The spatial position data of the marker points and the velocity-angle change data are fused according to a preset second weight ratio to obtain the enhanced end joint motion data of the tested object. The weight of the spatial position data of the marker points is less than the weight of the velocity-angle change data. Based on high-precision position data, the trajectory change data of the end joint of the tested object in space is obtained; Based on enhanced joint motion data, the velocity change data of the end joint of the tested object in space are obtained; By combining trajectory change data and velocity change data, the motion data of the target end joint of the tested object is output.

[0033] In application, the spatial position data of the marker points and the velocity-angle change data are first fused according to a second weighting ratio: 40% for marker point data and 60% for velocity-angle data. For example, the wrist marker point position (1.1365, 0.6345, 0.0785) and the velocity-angle data (velocity 0.4 m / s, angle 5°) are fused to obtain enhanced end-joint motion data: position (1.137, 0.635, 0.079), velocity 0.41 m / s, angle 5.1°. Then, based on the high-precision position data, trajectory change data is obtained: the athlete's wrist trajectory changes from (1.1315, 0.6315, 0.0785) to (1.1415, 0.6415, 0.0885) in 0.2 seconds, with a change rate of 0.05 m / s. Next, based on the enhanced joint motion data, velocity change data was obtained: the velocity increased from 0.4 m / s to 0.42 m / s. Finally, combining the trajectory change and velocity change, the target end joint motion data was output: a time series data packet, such as timestamp 10:00:00.055, containing the position (1.139, 0.637, 0.083), velocity 0.415 m / s, and angle 5.05°.

[0034] This invention provides a device for enhancing the end-joint accuracy of a markerless motion capture system, using any one of the methods described above for enhancing the end-joint accuracy of a markerless motion capture system. The device includes the following: Main system module 1: Used to acquire video images and extract key points of the human skeleton to obtain the raw motion capture data of the tested object. The main system module includes at least two cameras that acquire data simultaneously. Enhancement module 2: It is used to be installed at the end joint of the object under test and to collect the second end joint position data and enhanced end joint position data, wherein the end joints include the wrist, fingertips, ankle and toe tips. The enhancement module includes at least one optical marker and an inertial sensor. Fusion Module 3: This module is used to fuse the data output from the main system module and the enhancement module, and output the fused target end joint motion data.

[0035] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A method for end-joint precision enhancement of a markerless motion capture system, the method comprising: The method comprises the following steps: A main system collects original motion capture data of a measured object, performs data recognition and screening on the original motion capture data, and obtains a preliminary joint pose of the whole body of the measured object; An enhanced module is installed at one or more end joints of the measured object, and the enhanced module at least comprises an optical marker point and an inertial sensor; Based on the preliminary joint pose, spatial position recognition is performed on the end joints of the measured object to obtain first end joint position data, and second end joint position data of the enhanced module is synchronously collected; The first end joint position data and the second end joint position data are subjected to fusion filtering processing to obtain high-precision position data of the end joints; Enhanced end joint motion data of the enhanced module are collected, and target end joint motion data are output by combining the enhanced end joint motion data and the high-precision position data.

2. The end joint accuracy enhancement method of a markerless motion capture system of claim 1, wherein, The step of performing data recognition and screening on the original motion capture data to obtain a preliminary joint pose of the whole body of the measured object comprises the following steps: Based on the original motion capture data, joint recognition is performed on the original motion capture data to obtain joint positions of the measured object; According to the joint positions, data screening is performed on the original motion capture data to obtain initial joint motion data corresponding to each joint position; A spatial coordinate system is constructed, and joint motion trajectories and joint action changes are plotted according to the initial joint motion data in the spatial coordinate system; Based on the joint action changes, joint postures are obtained, based on the joint motion trajectories, joint spatial positions are obtained, and the joint preliminary pose is obtained by combining the joint postures and the joint spatial positions.

3. The end joint accuracy enhancement method of a markerless motion capture system of claim 2, wherein, The step of performing spatial position recognition on the end joints of the measured object based on the preliminary joint pose to obtain first end joint position data comprises the following steps: Based on the preliminary joint pose, the joint types corresponding to each preliminary joint pose are recognized, and target preliminary joint poses corresponding to the end joints are obtained by screening according to the joint types; Based on the target preliminary joint poses, the joint spatial positions in each target preliminary joint pose are extracted; The target joint types of the end joints are extracted, joint labels are generated according to the target joint types, the joint labels are added to the joint spatial positions, and first end joint position data are obtained.

4. The end joint accuracy enhancement method of a markerless motion capture system of claim 3, wherein, The step of collecting second end joint position data of the enhanced module comprises the following steps: Based on the optical marker point, a camera photographs the optical marker point to obtain marker point spatial position data of the optical marker point; Based on the inertial sensor, the inertial sensor records the acceleration and angular velocity of the end joint; Based on the acceleration and the angular velocity, velocity-angle change data of the end joint in the motion process are obtained; The marker point spatial position data and the velocity-angle change data are fused according to a preset first weight ratio to obtain second end joint position data of the end joint of the measured object, wherein the weight of the marker point spatial position data is greater than the weight of the velocity-angle change data.

5. A method of end joint precision enhancement for a markerless motion capture system according to claim 4, wherein, The method further comprises, before the step of collecting the second end joint position data of the enhancement module: extracting a data collection frequency of the main system during the collection of the original motion capture data; the enhancement module synchronously records data based on the data collection frequency and the main system.

6. A method of end joint precision enhancement for a markerless motion capture system according to claim 5, wherein, The step of fusing the first end joint position data and the second end joint position data to obtain high-precision position data of the end joint comprises: extracting a first time series data sequence of the first end joint position data based on the first end joint position data; extracting a second time series data sequence of the second end joint position data based on the second end joint position data; performing data clarity value evaluation on the first end joint position data to obtain a clarity value index of the first end joint position data; performing dynamic weight distribution on the first end joint position data and the second end joint position data according to the clarity value index to obtain a weight ratio; aligning the first time series data sequence and the second time series data sequence, and fusing the first end joint position data and the second end joint position data according to the weight ratio to obtain the high-precision position data of the end joint.

7. A method of end joint precision enhancement for a markerless motion capture system according to claim 6, wherein, The method further comprises, after the step of fusing the first end joint position data and the second end joint position data: extracting a full-body posture of the measured object in the original motion capture data, and recording fused data of the first end joint position data and the second end joint position data after fusion; matching the fused data with the full-body posture to obtain a matching value between the full-body posture of the measured object and the fused data; evaluating a local precision value of the fused data according to the matching value, and determining whether the local precision value is higher than a preset precision threshold; if it is determined that the local precision value is lower than the precision threshold, generating an accuracy error correction parameter based on the matching value, and correcting the fused data in accuracy according to the accuracy error correction parameter.

8. The method of end joint precision enhancement for a markerless motion capture system of claim 7, wherein, The step of collecting enhancement end joint motion data of the enhancement module, combining the enhancement end joint motion data and the high-precision position data, and outputting target end joint motion data comprises: fusing the marker point space position data and the speed-angle change data according to a preset second weight ratio to obtain enhancement end joint motion data of the measured object, wherein the weight of the marker point space position data is less than the weight of the speed-angle change data; obtaining trajectory change data of the end joint of the measured object in space based on the high-precision position data; obtaining speed change data of the end joint of the measured object in space based on the enhancement joint motion data; combining the trajectory change data and the speed change data to output target end joint motion data of the measured object.

9. An end joint precision enhancement apparatus for a markerless motion capture system, the apparatus using a markerless motion capture system end joint precision enhancement method as claimed in any one of claims 1 to 8, characterized by, The device comprises: The main system module is used for collecting video images and extracting human skeleton key points to obtain original motion capture data of the measured object, wherein the main system module comprises at least two cameras for synchronously collecting data; The enhancement module is used for being installed at the end joint of the measured object and collecting second end joint position data and enhanced end joint position data of the end joint, wherein the end joint comprises a wrist, a finger end, an ankle and a toe end, and the enhancement module comprises at least one optical marker point and an inertial sensor; The fusion module is used for fusing the data output by the main system module and the enhancement module and outputting target end joint motion data after fusion processing.

Images