A vibration-resistant multi-camera motion capture system and data correction method thereof

Abstract

The present invention relates to the field of data processing technology, and specifically to a vibration-resistant multi-group camera motion capture system and a data correction method thereof. The method comprises the following steps: collecting image data of a falling target by using multiple infrared motion capture cameras on multiple independent metal frame brackets; performing multi-perspective cross-validation processing on the image data collected by the multiple groups of cameras based on spatial reconstruction consistency, identifying and eliminating abnormal parallax data caused by camera vibration; performing trajectory filtering processing on the data after the cross-validation processing, and combining Kalman filtering with adaptive weighted median filtering to smooth the trajectory data; dividing the filtered trajectory data into multiple overlapping time windows according to time; and ensuring accurate trajectory tracking in high-dynamic impact scenarios, thereby improving the authenticity and reliability of the captured data.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a vibration-resistant multi-group camera motion capture system and a data correction method thereof. Background Art

[0002] Existing 3D motion capture systems typically use multiple infrared cameras arranged in a fixed array around the capture space, capturing the object's motion trajectory through triangulation. However, when used in high-impact scenarios, such as when an object free-falls from a high altitude and impacts the ground, strong vibrations can be transmitted to the camera's mount or foundation, causing small but significant relative displacements, thereby compromising the accuracy and consistency of the measurement data. While some high-end motion capture systems employ active calibration or high-frequency relocalization algorithms, these systems are still unable to fully eliminate tracking errors caused by relative shifts between cameras in environments subject to instantaneous and severe vibrations.

[0003] To this end, we propose a vibration-resistant multi-camera motion capture system and its data correction method to solve the above problems. Summary of the Invention

[0004] The object of the present invention is to provide a vibration-resistant multi-camera motion capture system and a data correction method thereof to solve the problems raised in the above background technology.

[0005] To achieve the above objectives, the present invention provides the following technical solutions: a vibration-resistant multi-camera motion capture system and a data correction method thereof, the method comprising the following steps:

[0006] Image data of falling targets are collected by using multiple infrared motion capture cameras on multiple independent metal frame brackets;

[0007] Based on the consistency of spatial reconstruction, the image data collected by multiple cameras are cross-validated from multiple perspectives to identify and eliminate abnormal parallax data caused by camera vibration.

[0008] Perform trajectory filtering on the data after cross-validation processing, and combine Kalman filtering with adaptive weighted median filtering to smooth the trajectory data;

[0009] The filtered trajectory data is divided into multiple overlapping time windows according to time, and the trajectory is reconstructed by performing local polynomial fitting in each time window;

[0010] The micro-displacement of the system at different vibration frequencies is mapped and modeled, and error compensation is iteratively optimized through Bayesian filtering.

[0011] Preferably, the specific steps of the multi-view cross-validation process are:

[0012] Extract the target's feature points from the collected image data, and use the triangulation principle to reconstruct the feature points observed by each capture camera into three-dimensional space to obtain the three-dimensional coordinates of each feature point. ; For the Feature points observed by the camera , calculate its reprojection error , the corresponding calculation formula is ,in, Indicates the The projection matrix of the camera.

[0013] Preferably, after calculating the reprojection error, the method further includes: setting a tolerance error threshold ,like , then the point is determined to be an abnormal point caused by vibration and is removed; the data after removing the abnormal point is used as the valid data for subsequent processing.

[0014] Preferably, the specific steps of the trajectory filtering process are:

[0015] Initialize the state vector ,in, Indicates the target location, Indicates the target speed. The initial state can be set based on the initial observation data or preset value.

[0016] Building a state transition model ,in, , is the state transition matrix, is the process noise;

[0017] Building an observation model , maps the state vector to the observation space, where For in time Observation data obtained from sensors at all times; represents the observation matrix; is the observation noise;

[0018] The Kalman filter algorithm is used to smooth the trajectory data to obtain the state estimation value ;

[0019] Apply a window on each dimension Amplitude median operation, remove outliers, the corresponding formula is ,in, is the window size of the median filter, Represents the observation value after median filtering;

[0020] The Kalman filter output and the median filter output are weighted and fused to obtain the final filtering result. The corresponding formula is: ,in, Control filter strength parameters, Represents the trajectory data after weighted fusion.

[0021] Preferably, the step of dividing the trajectory data after filtering into a plurality of overlapping time windows according to time, and performing local polynomial fitting in each time window to reconstruct the trajectory includes:

[0022] Divide the high-speed falling process into multiple overlapping time windows , the length of each time window is , the overlapping part is used for smooth transition, Indicates the start time of the time window;

[0023] In each time window, the trajectory data is reconstructed by local polynomial fitting, assuming that the target is at time The coordinate value of , the formula for fitting the second-order polynomial model is ,in, are polynomial fitting parameters, is the fitting residual;

[0024] The optimal parameter estimates are determined by minimizing the sum of squares of the fitted residuals. ; Use the fitting parameters to reconstruct the trajectory data in each time window, obtain the trajectory estimation value at each time point through the polynomial model, and reconstruct the motion trajectory in the time window;

[0025] During the window sliding process, each trajectory is weighted average fused, and the corresponding formula is: ,in, It is The trajectory fitted in the time window, is the corresponding weight, Represents the final trajectory after weighted average fusion.

[0026] Preferably, the steps of mapping and modeling the micro-displacement of the system at different vibration frequencies and iteratively optimizing error compensation through Bayesian filtering include:

[0027] The vibration behavior of each capture camera is modeled as a low-frequency jitter function, specifically expressed as ,in, For the Camera at all times The rotation error, For the Camera at all times The translation error is ; the small angle approximation is used to calculate the rotation error, and the corresponding formula is ,in, is the micro-rotation matrix, It is represented by an antisymmetric matrix; the vibration model is fitted by the drift of the calibration points before and after excitation. Using the known calibration point positions, measurements are taken before and after vibration to obtain displacement data, and the parameters of the vibration model are fitted through multiple displacement data; the fitted vibration model parameters are used as error feedback, and the error compensation is iteratively optimized through Bayesian filtering. The error estimate is continuously updated using prior knowledge and observation data.

[0028] Preferably, the vibration model is fitted by the drift of the calibration points before and after the excitation, and the displacement data are obtained by measuring the known calibration point positions before and after the vibration, and the parameters of the vibration model are fitted by the multiple displacement data.

[0029] Before vibration, record the position of the calibration point , after the vibration, record the position of the calibration point , calculate the displacement of the calibration point , using displacement data to fit vibration model parameters.

[0030] Preferably, the specific steps of using the vibration model parameters obtained by fitting as error feedback, iteratively optimizing error compensation through Bayesian filtering, and continuously updating the error estimate using prior knowledge and observation data to achieve online error back-estimation and correction include:

[0031] Initialize error estimate and the error covariance matrix , at each time step , based on new observational data Update the error estimate, the corresponding formula is ,in, is the Kalman gain, is the observation matrix;

[0032] The formula corresponding to the updated error covariance matrix is ,in, No. Camera at all times The vibration error, represents the identity matrix, is the error covariance matrix.

[0033] A vibration-resistant multi-group camera motion capture system includes multiple groups of independent brackets, and each group of brackets is equipped with multiple infrared motion capture cameras. The optical axes of the capture cameras are all directed to a central test area. A cubic falling target for testing is set in the central test area. The falling target has an infrared reflective mark. The camera is fixed to a metal frame through a multi-dimensional adjustment device. A spring-damping combination seismic isolation device is set at the bottom of the bracket to absorb ground vibration caused by the impact of the falling object. The image data collected by the capture system executes any of the data correction methods described above.

[0034] Compared with the prior art, the present invention has the following beneficial effects:

[0035] 1. Integrating multi-view cross-validation, trajectory filtering, time window reconstruction, and micro-jitter compensation algorithms, the collected data is processed and corrected layer by layer to avoid data errors caused by camera micro-displacement due to severe vibrations. This ensures accurate trajectory tracking in high-dynamic impact scenarios, improving the authenticity and reliability of captured data.

[0036] 2. Multiple infrared motion capture cameras mounted on independent metal frame brackets collect image data of falling targets from multiple perspectives, effectively avoiding the limitations of a single perspective and significantly improving the comprehensiveness and accuracy of the data. This ensures the authenticity of the motion capture data and the stability of the camera spatial layout under ground vibration. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0038] Figure 1 Schematic diagram of the method flow of the present invention;

[0039] Figure 2 This is a system structure diagram of the present invention. DETAILED DESCRIPTION

[0040] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0041] Example

[0042] See also Figures 1 to 2 The present invention provides a technical solution for a vibration-resistant multi-group camera motion capture system and a data correction method thereof: A data correction method for a vibration-resistant multi-group camera motion capture system comprises the following steps:

[0043] S1: Collect image data of falling targets through multiple infrared motion capture cameras on multiple independent metal frame brackets;

[0044] S2: Based on the consistency of spatial reconstruction, the image data collected by multiple cameras are cross-validated from multiple perspectives to identify and eliminate abnormal parallax data caused by camera shake;

[0045] The specific steps of the multi-view cross-validation process are: extracting the target's feature points from the collected image data, such as the position of the infrared reflective marker; using the triangulation principle, reconstructing the feature points observed by each capture camera into three-dimensional space to obtain the three-dimensional coordinates of each feature point ; For the Feature points observed by the camera , calculate its reprojection error The corresponding calculation formula is ,in, Indicates the The projection matrix of the camera is used to project the three-dimensional space point onto the two-dimensional image plane; set a tolerance error threshold ,like , then the point is determined to be an abnormal point caused by vibration and is removed; the data after the abnormal point is removed is used as the valid data for subsequent processing;

[0046] S3: Perform trajectory filtering on the data after cross-validation processing, combining Kalman filtering and adaptive weighted median filtering to smooth the trajectory data;

[0047] The specific steps of the trajectory filtering process are: initializing the state vector ,in, Indicates the target location, Represents the target speed. The initial state can be set based on the initial observation data or preset value; build a state transition model , describes the change relationship of the state vector over time, where , is the state transfer matrix, usually a linear matrix, which represents the state vector at the time step arrive The changing relationship is process noise, which indicates the uncertainty of the system model and is usually assumed to be zero-mean Gaussian white noise; construct the observation model , maps the state vector to the observation space, where is the observed value, indicating that Observation data obtained from sensors at all times; Represents the observation matrix, which is used to map the state vector to the observation space; is the observation noise, which represents the random error in the observation process, and is usually assumed to be zero-mean Gaussian white noise; the trajectory data is smoothed using the Kalman filter algorithm to obtain the state estimate , Kalman filter achieves the optimal estimation of the target state by recursively updating the state estimate and error covariance matrix; applying a window on each dimension Amplitude median operation, remove outliers, the corresponding formula is ,in, is the window size of the median filter, which is used to determine the range of the median filter. The median filter effectively removes outliers and noise by taking the median of the data in the window. is the median filter output, which represents the observation value after median filtering. The Kalman filter output and the median filter output are weighted and fused to obtain the final filtering result. The corresponding formula is ,in, Control the filter strength parameter to adjust the weights of Kalman filter and median filter. is the final filtering result, which represents the trajectory data after weighted fusion. By combining the smoothing characteristics of the Kalman filter and the noise resistance of the median filter, the robustness of the trajectory data is improved. It can effectively smooth the trajectory data, remove noise and outliers, and improve the robustness and accuracy of the motion capture data.

[0048] S4: Divide the filtered trajectory data into multiple overlapping time windows according to time, and perform local polynomial fitting in each time window to reconstruct the trajectory to improve the robustness and accuracy of the reconstruction;

[0049] The steps of dividing the trajectory data after filtering into multiple overlapping time windows according to time and performing local polynomial fitting in each time window to reconstruct the trajectory include: dividing the high-speed falling process into multiple overlapping time windows , the length of each time window is , the overlapping part is used for smooth transition, Indicates the start time of the time window and the starting moment of the time window; the division of the time window can be adjusted according to the characteristics of the motion and the data sampling rate to ensure that there are enough data points in each time window for modeling; in each time window, the trajectory data is reconstructed by local polynomial fitting, and the target is set at time The coordinate value of , the formula for fitting a second-order polynomial model is ,in, is the polynomial fitting parameter, which is used to describe the changing trend of the trajectory within the time window. The fitting residual represents the difference between the actual trajectory and the fitted trajectory; the fitting parameters are estimated by the least squares method to minimize the fitting error. The least squares method finds the optimal parameter estimate by minimizing the sum of squares of the fitting residuals. ; Use fitting parameters to reconstruct the trajectory data in each time window; obtain the trajectory estimation value of each time point through the polynomial model, and thus reconstruct the motion trajectory in the time window; during the window sliding process, perform weighted average fusion on each trajectory to improve anti-shake performance and fitting accuracy. Weighted average fusion can use the data of the overlapping part of the time window to fuse by weighted summation. The corresponding formula is ,in, It is The trajectory fitted in the time window, is the corresponding weight, which can be adjusted according to the degree of overlap of the time window and the reliability of the data. is the fused trajectory, which represents the final trajectory after weighted average fusion. It can effectively improve the robustness and accuracy of trajectory data, especially in high-dynamic impact scenarios, and can significantly reduce the impact of vibration on trajectory reconstruction.

[0050] S5: Map and model the system's micro-displacement at different vibration frequencies, and iteratively optimize error compensation through Bayesian filtering to achieve online error back-calculation and correction;

[0051] The steps of mapping and modeling the micro-displacement of the system at different vibration frequencies and iteratively optimizing the error compensation through Bayesian filtering include: modeling the vibration behavior of each capture camera as a low-frequency jitter function, specifically expressed as ,in, For the Camera at all times The rotation error, For the Camera at all times The translation error reflects the slight displacement and rotation of the camera under vibration conditions; the small angle approximation is used to simplify the calculation of the rotation error. The corresponding formula is ,in, is the micro-rotation matrix, which represents the small-angle rotation deviation of the camera. It is an antisymmetric matrix representation used to convert vectors into antisymmetric matrices. The vibration model is fitted by the drift of calibration points before and after excitation. Using the known calibration point positions, measurements are taken before and after vibration to obtain displacement data. The vibration model parameters are fitted using multiple displacement data. The fitted vibration model parameters are used as error feedback, and error compensation is iteratively optimized through Bayesian filtering. The error estimate is continuously updated using prior knowledge and observation data, thereby achieving online error back-calculation and correction.

[0052] The vibration model is fitted by the drift of the calibration points before and after the excitation. Using the known calibration point positions, measurements are taken before and after the vibration to obtain displacement data. The specific steps for fitting the parameters of the vibration model through multiple displacement data are as follows: Before the vibration, record the position of the calibration points , after the vibration, record the position of the calibration point , calculate the displacement of the calibration point , using displacement data to fit vibration model parameters.

[0053] The vibration model parameters obtained by fitting are used as error feedback, error compensation is optimized iteratively through Bayesian filtering, and the error estimation is continuously updated using prior knowledge and observation data, thereby achieving online error back-calculation and correction. The specific steps include: initializing the error estimation and the error covariance matrix , at each time step , based on new observational data Update the error estimate, the corresponding formula is ,in, is the Kalman gain, is the observation matrix; the formula corresponding to the updated error covariance matrix is ,in, No. Camera at all times The vibration error indicates the position and attitude deviation of the camera. represents the identity matrix, represents the identity element of the matrix, The error covariance matrix represents the uncertainty of the error estimate and can effectively compensate for small displacement and rotation errors of the camera under vibration conditions, improving the accuracy and stability of motion capture data. This method is particularly suitable for high-dynamic impact scenarios such as drop tests, crash tests, and high-speed target tracking, significantly reducing the impact of vibration on trajectory reconstruction.

[0054] When the capture system collects data, all cameras perform synchronized exposures. A cross-validation module detects and removes data points with abnormal spatial positions. A filtering module then removes noise and smoothes the target trajectory. During the period of violent ground contact in free fall, the trajectory is reconstructed using a motion model within the time window. Any errors caused by slight camera offsets are automatically identified and compensated by the micro-jitter model.

[0055] A vibration-resistant multi-group camera motion capture system includes multiple groups of independent brackets, and each group of brackets is equipped with multiple infrared motion capture cameras. The optical axes of the capture cameras are all directed to a central test area. A cubic falling target for testing is set in the central test area. The falling target has an infrared reflective mark. The camera is fixed to a metal frame through a multi-dimensional adjustment device. A spring-damping combination seismic isolation device is set at the bottom of the bracket to absorb ground vibration caused by the impact of the falling object. The image data collected by the capture system executes any of the data correction methods described above.

[0056] Specifically, multiple groups of independent brackets refer to multiple groups of metal frame brackets, each group of brackets is independently supported and has no physical connection with each other, and is arranged on the ground; multiple infrared motion capture cameras are installed on each group of brackets, and the cameras are fixed to the metal frame through a multi-dimensional adjustment device, and the optical axis points to the central test area; a spring-damping combined isolation device is provided at the bottom of each bracket to absorb the ground vibration caused by the impact of the falling object; a cube falling target for testing is set in the central test area, and the target is marked with an infrared reflective marker. The isolation device includes a spring-damper combination device to absorb the ground vibration caused by the impact of the falling object and reduce the degree of vibration transmission to the camera bracket, thereby reducing the slight displacement of the camera caused by the vibration and ensuring the stability of the camera in high-dynamic impact scenarios; the cube falling target is a cube with an infrared reflective marker. Used for free fall testing, its infrared reflective markers can be accurately captured by infrared motion capture cameras, which can perform synchronous exposure to ensure that all cameras capture images of the target object at the same time, provide synchronized raw data for data correction, and improve the accuracy and consistency of trajectory reconstruction; multiple groups of metal frame brackets are arranged in a triangle, and the three metal frames are respectively recorded as brackets A, B, and C. Each bracket is installed with three cameras, namely CAM1-CAM9, facing the central falling area to ensure that the target is in multi-perspective coverage, thereby improving the redundancy of the system and the reliability of the data. Through three-point independent support, anti-seismic device, perspective redundancy design and data compensation algorithm, the system can still stably track the trajectory of the target object under vibration conditions such as ground impact and laboratory drop, realizing the coordinated optimization of structural stability and algorithm robustness.

[0057] Throughout this specification, references to terms such as "one embodiment," "example," or "specific example" indicate that the specific features, structures, materials, or characteristics described in conjunction with that embodiment or example are included in at least one embodiment or example of the present invention. In this specification, schematic representations of these terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0058] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A data correction method for a vibration-resistant multi-camera motion capture system, characterized in that: The following steps are involved: Image data of falling targets are collected by using multiple infrared motion capture cameras on multiple independent metal frame brackets; Based on the consistency of spatial reconstruction, the image data collected by multiple cameras are cross-validated from multiple perspectives to identify and eliminate abnormal parallax data caused by camera vibration. Perform trajectory filtering on the data after cross-validation processing, and combine Kalman filtering with adaptive weighted median filtering to smooth the trajectory data; The specific steps of the trajectory filtering process are: Initialize the state vector ,in, Indicates the target location, Indicates the target speed, the initial state is set based on the initial observation data or preset value; Building a state transition model ,in, , is the state transition matrix, is the process noise; Building an observation model , maps the state vector to the observation space, where For in time Observation data obtained from sensors at all times; represents the observation matrix; is the observation noise; The Kalman filter algorithm is used to smooth the trajectory data to obtain the state estimation value ; Apply a window on each dimension Amplitude median operation, remove outliers, the corresponding formula is ,in, is the window size of the median filter, Represents the observation value after median filtering; The Kalman filter output and the median filter output are weighted and fused to obtain the final filtering result. The corresponding formula is: ,in, Control filter strength parameters, Represents the trajectory data after weighted fusion; The filtered trajectory data is divided into multiple overlapping time windows according to time, and the trajectory is reconstructed by performing local polynomial fitting in each time window; The step of dividing the trajectory data after filtering into multiple overlapping time windows according to time, and performing local polynomial fitting in each time window to reconstruct the trajectory includes: Divide the high-speed falling process into multiple overlapping time windows , the length of each time window is , the overlapping part is used for smooth transition, Indicates the start time of the time window; In each time window, the trajectory data is reconstructed by local polynomial fitting, assuming that the target is at time The coordinate value of , the formula for fitting the second-order polynomial model is ,in, are polynomial fitting parameters, is the fitting residual; The optimal parameter estimates are determined by minimizing the sum of squares of the fitted residuals. ; Use the fitting parameters to reconstruct the trajectory data in each time window, obtain the trajectory estimation value at each time point through the polynomial model, and reconstruct the motion trajectory in the time window; During the window sliding process, each trajectory is weighted average fused, and the corresponding formula is: ,in, It is The trajectory fitted in the time window, is the corresponding weight, represents the final trajectory after weighted average fusion; The micro-displacement of the system at different vibration frequencies is mapped and modeled, and error compensation is iteratively optimized through Bayesian filtering.

2. The data correction method for a vibration-resistant multi-camera motion capture system according to claim 1, characterized in that: The specific steps of the multi-view cross-validation process are: Extract the target's feature points from the collected image data, and use the triangulation principle to reconstruct the feature points observed by each capture camera into three-dimensional space to obtain the three-dimensional coordinates of each feature point. ; For the Feature points observed by a camera , calculate its reprojection error , the corresponding calculation formula is ,in, Indicates the The projection matrix of the camera.

3. The data correction method for a vibration-resistant multi-camera motion capture system according to claim 2, characterized in that: After calculating the reprojection error, it also includes: setting a tolerance error threshold ,like , then the point is determined to be an abnormal point caused by vibration and is removed; the data after removing the abnormal point is used as the valid data for subsequent processing.

4. The data correction method for a vibration-resistant multi-camera motion capture system according to claim 1, characterized in that: The steps of mapping and modeling the micro-displacement of the system at different vibration frequencies and iteratively optimizing error compensation through Bayesian filtering include: The vibration behavior of each capture camera is modeled as a low-frequency jitter function, specifically expressed as ,in, For the Camera at all times The rotation error, For the Camera at all times The translation error is ; the small angle approximation is used to calculate the rotation error, and the corresponding formula is ,in, is the micro-rotation matrix, It is represented by an antisymmetric matrix; the vibration model is fitted by the drift of the calibration points before and after excitation. Using the known calibration point positions, measurements are taken before and after vibration to obtain displacement data, and the parameters of the vibration model are fitted through multiple displacement data; the fitted vibration model parameters are used as error feedback, and the error compensation is iteratively optimized through Bayesian filtering. The error estimate is continuously updated using prior knowledge and observation data.

5. The data correction method for a vibration-resistant multi-camera motion capture system according to claim 4, characterized in that: The vibration model is fitted by the drift of the calibration points before and after the excitation, and the displacement data are obtained by measuring the known calibration point positions before and after the vibration, and the parameters of the vibration model are fitted by the multiple displacement data. Before vibration, record the position of the calibration point , after the vibration, record the position of the calibration point , calculate the displacement of the calibration point , using displacement data to fit vibration model parameters.

6. The data correction method for a vibration-resistant multi-camera motion capture system according to claim 4, characterized in that: The specific steps of using the fitted vibration model parameters as error feedback, iteratively optimizing error compensation through Bayesian filtering, and continuously updating the error estimate using prior knowledge and observation data to achieve online error back-estimation and correction include: Initialize error estimate and the error covariance matrix , at each time step , based on new observational data Update the error estimate, the corresponding formula is ,in, is the Kalman gain, is the observation matrix; The formula corresponding to the updated error covariance matrix is ,in, No. Camera at all times The vibration error, represents the identity matrix, is the error covariance matrix.

7. A vibration-resistant multi-camera motion capture system, characterized in that: The invention comprises a plurality of independent brackets, and each bracket is provided with a plurality of infrared motion capture cameras, the optical axes of the capture cameras are all directed to the central test area, a cubic falling target for the test is provided in the central test area, and the falling target has an infrared reflective mark, the camera is fixed on the metal frame through a multi-dimensional adjustment device, and a spring-damping combination seismic isolation device is provided at the bottom of the bracket to absorb the ground vibration caused by the impact of the falling object, and the image data collected by the capture system executes the data correction method described in any one of claims 1 to 6.

Images