Method for measuring spatial precision of optical motion capture system
Through the joint acquisition and data processing of laser tracker and optical motion capture system, the problem of uneven measurement accuracy of optical motion capture system in large-scale space is solved, and the effective evaluation and improvement of the spatial accuracy of optical motion capture system is achieved.
Patent Information
- Application Number
- CN202411624633.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-14
- Publication Date
- 2025-02-28
Smart Images
Figure CN119533446A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the field of spatial accuracy calculation, and in particular to a method for measuring the spatial accuracy of an optical motion capture system. Background Art
[0002] The optical motion capture system uses an infrared camera to collect infrared signals reflected from a passive reflective ball to locate objects. However, the measurement accuracy of optical motion capture technology is affected by many factors, such as calibration errors of the optical system, changes in ambient lighting conditions, reflective characteristics of the target object surface, the number and distribution of cameras, etc. These factors may lead to deviations in the measurement data, affect the distribution of measurement accuracy within the measurement space, and thus affect the accuracy and reliability of subsequent analysis. Especially in a larger range of detection space (such as more than 5m), the quality of illumination is difficult to guarantee, and the increase in measurement distance will also cause a decrease in imaging resolution and image signal-to-noise ratio, resulting in insufficient imaging details of the object to be measured and increased geometric deformation of the image, which in turn leads to changes in positioning accuracy, resulting in uneven detection accuracy in the three-dimensional space of its detection, and it is difficult to fully obtain the internal measurement accuracy distribution characteristics in a large-scale space.
[0003] If the measurement space increases and the number of cameras increases, the distribution of measurement accuracy will become more complicated. Therefore, it is necessary to evaluate the spatial measurement accuracy of the optical motion capture system.
[0004] However, there is no method for evaluating the detection accuracy within the detection area. This also restricts the application of optical motion capture systems in the measurement field due to the inability to accurately measure. Summary of the invention
[0005] Purpose of the invention: To provide a method for measuring the spatial accuracy of an optical motion capture system to solve the above-mentioned problems existing in the prior art.
[0006] Technical solution: A method for measuring the spatial accuracy of an optical motion capture system, including an optical motion capture system and a laser tracker; the optical motion capture system is flexibly constructed by multiple infrared cameras; the T-mac end of the laser tracker is installed at the end of a mechanical arm, and the optical mark point is installed at the top of the T-mark end. The mechanical arm equipped with the T-mac end of the laser tracker moves and adjusts according to the planned path, that is, the set spatial three-dimensional test grid, and pauses and collects data according to the planned test points. During the pause of the set mechanical arm, the laser tracker collects the position of the T-mac end placed on the set mechanical arm, and the optical motion capture system synchronously collects and identifies the position of the optical mark point placed on the set mechanical arm, and the optical motion capture system and the laser tracker independently record their respective timestamps. Finally, the data collected by the laser tracker is used as a reference system, and the collected data of the optical tracking system and the collected data of the tracker are uniformly converted into coordinate systems through the Rodrigues equation, and the accuracy measurement results of the optical motion capture system are summarized.
[0007] In a further embodiment, the distance that the robot moves each time along the grid direction of the three-dimensional test grid in the detection space is 100 mm, including 84 pauses, each pause is 3 seconds. The shooting frequency of the optical motion capture system is 100 frames per second, and the robot acquires 300 points of data during the 3-second pause to form a local point cloud, which is the position of the test point. The position of each test point is the position of the optical mark point that can be identified by the optical motion capture system; the laser tracker synchronously collects coordinates once during the Mark point stop process, and the optical motion capture system and the laser tracker collect data simultaneously, and each independently records its own timestamp.
[0008] In a further embodiment, the optical motion capture system data processing includes four stages, specifically a preprocessing stage based on a motion state model, a scale error correction stage based on laser tracking data and high-confidence motion capture data, a representative point output stage of distribution characteristics, and a data comparison calculation stage.
[0009] In a further embodiment, in the preprocessing stage, the measurement sequences collected by the optical motion capture system and the laser tracker are accurately matched according to the recorded timestamps, and finally the data collected by the laser tracker is used as a reference system to measure the error through the root mean square of all samples.
[0010] In a further embodiment, the timestamp information of the point cloud data can be used to divide the collected point cloud data into two stages based on the time period: motion start and stop and motion stability. The data distribution of continuous time points is simulated by Gaussian process, and the time dependency is simulated by Gaussian function. The form of Gaussian function is as follows:
[0011]
[0012] Where t is time, which indicates the timestamp of the point cloud data; μ is the mean of the Gaussian function, which indicates the central time point of the stable phase of the motion; σ is the standard deviation of the Gaussian function, which controls the width of the Gaussian function;
[0013] The distribution of the Gaussian function on the time axis can be controlled by adjusting the values of μ and σ; μ is set to the midpoint of the entire time period, and σ is adjusted according to the specific situation.
[0014] In a further embodiment, the stability of the data measured by the optical motion capture system during each pause can be expressed as obtained by calculating the variance distribution of each point in each local point cloud on the three coordinates of the optical motion capture system measurement coordinate system; the obtained variance data reflects the degree of data dispersion of each point on the x, y and z axes. The smaller the dispersion, the more stable the measurement system, and the formula is as follows:
[0015]
[0016] in: is the variance of the X-axis point coordinates in the nth local point cloud, x ni is the X-axis coordinate value of the i-th point in the n-th local point cloud, It is the average value of the X-axis coordinate value of the point coordinates in the nth local point cloud.
[0017] In a further embodiment, each group of pause points collected by the optical motion capture system is filtered and output with representative points to obtain k representative points. The point set obtained by laser tracking data collection is After the two point sets are processed through registration, the root mean square error (RMSE) of the Euclidean distance between the aligned points is used to measure the acquisition accuracy of the optical motion capture system;
[0018] The distance error DE of the i-th observation point is:
[0019]
[0020] Overall root mean square error RMSE:
[0021]
[0022] Where: k is the number of pause points, is the measured coordinate of the optical motion capture system at the ith stop point, is the measured coordinate of the laser tracking point at the i-th stop point.
[0023] In a further embodiment, the point set collected when the robot arm stops is clustered by distance clustering, and the marked points are extracted into different subsets, and the center of each cluster is used as the representative point of the cluster; the mean of all cluster centers is weightedly calculated according to the number of data points in the cluster;
[0024] For each cluster Ci, the mean of all points is calculated as the cluster center:
[0025]
[0026] Where p is a point in cluster Ci, |C i. size| is the number of data points in cluster Ci.
[0027] In a further embodiment, the weighted mean of all cluster centers is calculated by using the cluster centers and the number of data points of each cluster center, taking the number of data points as a weight:
[0028]
[0029] Where i is the index of the cluster, C i. size is the cluster C i The number of data points in C i Center is cluster C i center.
[0030] Beneficial effects: The present invention relates to a method for measuring the spatial accuracy of an optical motion capture system. The present invention collects spatial grid data simultaneously through a laser tracker and an optical motion capture system, and processes and analyzes the collected data through a corresponding algorithm: (1) Optimize the point cloud data in the path through a radius filtering algorithm; (2) Obtain effective spatial measurement point data through a proportional error correction algorithm; (3) Align the two sets of data coordinates through coordinate conversion so that the two sets of data can be compared under the same coordinates. (4) Effectively compare and analyze the two sets of data, thereby evaluating the spatial data accuracy of the optical motion capture system. The evaluation method enables the spatial accuracy of the optical motion capture system to be effectively evaluated, and promotes the application and development of the optical motion capture system in the field of spatial position and posture measurement. BRIEF DESCRIPTION OF THE DRAWINGS
[0031] Figure 1 Schematic diagram of a tower construction scene of the optical motion capture system of the present invention.
[0032] Figure 2This is a schematic diagram of the spatial three-dimensional test grid described in the present invention.
[0033] Figure 3 This is a schematic diagram of the installation method of the identification point of the present invention.
[0034] Figure 4 It is a schematic flow chart of the accuracy evaluation method of the optical motion capture system described in the present invention. DETAILED DESCRIPTION
[0035] In the following description, a large number of specific details are provided to provide a more thorough understanding of the present invention. However, it is apparent to those skilled in the art that the present invention can be implemented without one or more of these details. In other examples, in order to avoid confusion with the present invention, some technical features well known in the art are not described.
[0036] The applicant believes that as the measurement space increases and the number of cameras increases, the distribution of measurement accuracy becomes more complicated. Therefore, it is necessary to evaluate the spatial measurement accuracy of the optical motion capture system.
[0037] To this end, the applicant designed a method for measuring the spatial accuracy of an optical motion capture system. The applicant optimized and corrected the data collected by the optical motion capture system through a laser tracker, and evaluated the spatial data accuracy of the optical motion capture system through comparative analysis to improve the spatial data accuracy of the optical motion capture system.
[0038] The method for measuring the spatial accuracy of an optical motion capture system involved in the present invention mainly includes an optical motion capture system and a laser tracker; the optical motion capture system is flexibly constructed by multiple infrared cameras; the T-mac end of the laser tracker is installed at the end of a mechanical arm, and an optical marking point is installed at the top of the T-mark end. The mechanical arm equipped with the T-mac end of the laser tracker moves and adjusts according to the planned path, that is, the set spatial three-dimensional test grid, and pauses and collects data according to the planned test points. At the same time, the optical motion capture system synchronously collects and marks the laser tracker during the pause period, and the optical motion capture system and the laser tracker independently record their respective timestamps. Finally, the data collected by the laser tracker is used as a reference system, and the collected data of the optical tracking system and the collected data of the tracker are uniformly converted into coordinate systems through the Rodrigues equation, and the accuracy measurement result of the optical motion capture system is summarized.
[0039] In order to evaluate the accuracy of the optical motion capture system in optical space measurement, the present invention designs a spatial three-dimensional test grid, such as Figure 2As shown in FIG. 1 , in order to obtain a controllable three-dimensional test grid, the present invention performs three-dimensional spatial path planning on the industrial robot arm. The measuring instrument used in this method is a laser tracker, and the measurement accuracy of the laser tracker is an order of magnitude higher than that of the optical motion capture system, which meets the measurement conditions. In the measurement process, Figure 3 As shown, the T-mac end of the laser tracker is installed at the end of the robot arm, and the optical marking point is installed at the top of the T-mark end.
[0040] The distance that the robot arm moves each time along the grid direction of the set three-dimensional test grid in the detection space is 100 mm, including 84 pauses, each pause is 3 seconds. The shooting frequency of the optical motion capture system is 100 frames per second. The robot arm acquires 300 points of data during the 3-second pause to form a local point cloud, which is the position of the test point. The position of each test point is the position of the optical mark point that can be identified by the optical motion capture system. The laser tracker synchronously collects coordinates once during the Mark point stop process. The optical motion capture system and the laser tracker collect data simultaneously, and each independently records its own timestamp.
[0041] The data processing of the optical motion capture system includes four stages, specifically a preprocessing stage based on a motion state model, a scale error correction stage based on laser tracking data and high-confidence motion capture data, a representative point output stage of distribution characteristics, and a data comparison calculation stage.
[0042] In the preprocessing stage, the measurement sequences collected by the optical motion capture system and the laser tracker are accurately matched according to the recorded timestamps, and finally the data collected by the laser tracker is used as the reference system to measure the error through the root mean square of all samples.
[0043] The timestamp information of the point cloud data can be used to divide the collected point cloud data into two stages based on the time period: motion start and stop, and motion stability. The data distribution of continuous time points can be simulated by Gaussian process, and the time dependency can be simulated by Gaussian function. The form of Gaussian function is as follows:
[0044]
[0045] Where t is time, which indicates the timestamp of the point cloud data; μ is the mean of the Gaussian function, which indicates the central time point of the stable phase of the motion; σ is the standard deviation of the Gaussian function, which controls the width of the Gaussian function;
[0046] The distribution of the Gaussian function on the time axis can be controlled by adjusting the values of μ and σ; set μ to the midpoint of the entire time period, and adjust σ according to the specific situation. The value of the Gaussian function will be lower during the start and stop phase of the motion, and higher during the stable phase of the motion. In this way, by using the value of the Gaussian function as a weight, the weight of the point cloud data can be adjusted dynamically in time. Apply the calculated time-dynamic weight to the point cloud data. By multiplying the weight, the contribution of each point is adjusted so that the impact on the final result is smaller during the start and stop phase of the motion.
[0047] The data measurement equipment uses the laser tracker Leica AT960, including the host computer, controller, and six-dimensional posture tracking target T-MAC. It has a dual laser system of laser interference and absolute ranging, which maintains higher accuracy and stability on the basis of a larger range. The position measurement accuracy is 15um+6um / m. The T-Mac of the Leica laser tracker is a standard six-dimensional posture component, which realizes the real-time dynamic acquisition of 1000HZ six-dimensional data through multi-point monitoring and tracking functions. It has rigorous capability indicators and can be traced and verified through international standards. At the same time, it has an open software development port, which can be developed based on general operation commands. The experimental design uses a laser tracker to measure the posture of the end of the robotic arm. The measurement system consists of two parts: the laser tracker host and the six-dimensional posture tracking target. During the measurement process, the measurement laser of the laser tracker only needs to aim at the corner cube prism to realize the collection of posture data, and the measurement efficiency is high. In the experiment, the six-dimensional posture tracking target is installed at the end of the robotic arm, and three Mark points are installed on the upper end of the six-dimensional posture tracking target to construct the six-dimensional posture tracking target. Optical tracking and laser tracker three-dimensional dynamic combined measurement system.
[0048] From the measurement principles of the laser tracker and the optical tracking system, we can see that the data measured by the two are in different forms. The former is the single-point coordinate measured by T-Mac (the six-dimensional posture tracking target at the stop point), and the latter is the coordinate obtained by the optical camera to establish the center of the rigid body. The frequency is 0.01 seconds per frame, that is, 300 coordinate points will be obtained at each stop point. Moreover, the coordinate data measured by the two are not in the same coordinate system, and the coordinate origin is different, so the accuracy of the optical tracking system measurement data cannot be directly evaluated. First, the data measured by the optical tracking system needs to be aligned. The method used in this paper to align the data of the two is to use the Rodriguez rotation equation to rotate and translate the matrix.
[0049] R=I+(sinθ)K+(1-cosθ)K 2
[0050] Through computer algorithm language, the coordinate data of the center point of the rigid body measured in the optical tracking system is converted into the laser tracker coordinate system using the Rodriguez rotation equation and translation formula.
[0051] The stability of the data measured by the optical motion capture system during each pause can be expressed as the variance distribution of each point in each local point cloud on the three coordinates of the optical motion capture system measurement coordinate system; the obtained variance data reflects the degree of data dispersion of each point on the x, y and z axes. The smaller the dispersion, the more stable the measurement system. The formula is as follows:
[0052]
[0053] in: is the variance of the X-axis point coordinates in the nth local point cloud, x ni is the X-axis coordinate value of the i-th point in the n-th local point cloud, is the average value of the X-axis coordinate value of the point coordinate in the nth local point cloud. In order to evaluate the spatial measurement accuracy, the optical tracking system measurement value was compared with the laser tracker measurement value. The error distribution evaluation was also performed.
[0054] Each group of pause points collected by the optical motion capture system is filtered and output with representative points to obtain k representative points. The point set obtained by laser tracking data collection is After the two point sets are processed through registration, the root mean square error (RMSE) of the Euclidean distance between the aligned points is used to measure the acquisition accuracy of the optical motion capture system;
[0055] The distance error DE of the i-th observation point is:
[0056]
[0057] Overall root mean square error RMSE:
[0058]
[0059] Where: k is the number of pause points, is the measured coordinate of the optical motion capture system at the ith stop point, is the measured coordinate of the laser tracking point at the i-th stop point.
[0060] The point set collected when the robot arm stops is clustered by distance clustering, and the marked points are extracted into different subsets, and the center of each cluster is used as the representative point of the cluster; the mean of all cluster centers is weightedly calculated according to the number of data points in the cluster;
[0061] For each cluster Ci, the mean of all points is calculated as the cluster center:
[0062]
[0063] Where p is a point in cluster Ci, |C i. size| is the number of data points in cluster Ci.
[0064] By using the cluster centers and the number of data points in each cluster, the weighted mean of all cluster centers is calculated using the number of data points as the weight:
[0065]
[0066] Where i is the index of the cluster, C i. size is the cluster C i The number of data points in C i· Center is cluster C i center.
[0067] In a further preferred embodiment, during the pause of the marking point, the optical motion tracking system continuously observes the marking point for about n=300 times, using the distance variance The distance standard deviation σd is used to quantify the discreteness of the set of observations. The coordinates of the stop position identification points obtained each time are (x1, y1, z1), (x2, y2, z2), …, (xn, yn, zn), and the mean of these observations, that is, the center point is (xc, yc, zc), where:
[0068]
[0069] The distance between each observed point and the center point is di:
[0070]
[0071] The variance of these distances formula:
[0072]
[0073] Where μd is the average of all distances di:
[0074]
[0075] The error ellipsoid can analyze the distribution and dispersion of the observations of the motion tracking system in three-dimensional space. Its shape and size are determined by the covariance matrix of the observations. For a set of three-dimensional coordinates (xi, yi, zi) obtained from n observations during a pause, where i = 1, 2, ..., n. First, the center point (xc, yc, zc) of the observations is obtained according to formula (1). Then, the covariance matrix C of the observations is calculated. The covariance matrix is a 3×3 matrix whose element Cij represents the covariance between the i-th coordinate component and the j-th coordinate component.
[0076]
[0077] Among them, Var(x), Var(y), Var(z) are the variances of the x, y, and z coordinate components respectively, and Cov(x,y), Cov(x,z), Cov(y,z) are the corresponding covariances. These values can be calculated by the following formula:
[0078]
[0079]
[0080] The shape and size of the error ellipsoid can be described by the covariance matrix. The direction and length of the principal axis of the error ellipsoid are determined by the eigenvectors and eigenvalues of the covariance matrix.
[0081] As described above, although the present invention has been shown and described with reference to specific preferred embodiments, it should not be construed as limiting the present invention itself. Various changes in form and details may be made without departing from the spirit and scope of the present invention as defined in the appended claims.
Claims
1. A method for measuring the spatial accuracy of an optical motion capture system, characterized by: It includes an optical motion capture system and a laser tracker; the optical motion capture system is flexibly constructed by multiple infrared cameras; the T-mac end of the laser tracker is installed at the end of the robotic arm, and the optical marking point is installed at the top of the T-mark end; The robotic arm equipped with the T-mac end of the laser tracker moves and adjusts according to the planned path, that is, the set spatial three-dimensional test grid, and pauses and collects data according to the planned test points. During the pause of the set robotic arm, the laser tracker collects the position of the T-mac end placed on the set robotic arm, and the optical motion capture system synchronously collects and marks the position of the optical marker point placed on the set robotic arm, and the optical motion capture system and the laser tracker independently record their respective timestamps. Finally, the data collected by the laser tracker is used as a reference system, and the collected data of the optical tracking system and the collected data of the tracker are uniformly converted into coordinate systems through the Rodrigues equation, and the accuracy measurement results of the optical motion capture system are summarized.
2. The method for measuring the spatial accuracy of an optical motion capture system according to claim 1, characterized in that: The distance each movement of the robot arm along the grid direction of the set three-dimensional test grid in the detection space is 100 mm, including 84 pauses, each pause is 3 seconds; The shooting frequency of the optical motion capture system is 100 frames per second. The data of 300 points are acquired during the 3-second pause of the robotic arm to form a local point cloud, which is the position of the test point. The position of each test point is the position of the pause of the optical marker point identifiable by the optical motion capture system; the laser tracker synchronously collects coordinates once during the stay of the Mark point. The optical motion capture system and the laser tracker collect data simultaneously, and each independently records its own timestamp.
3. The method for measuring the spatial accuracy of an optical motion capture system according to claim 2, characterized in that: The data processing of the optical motion capture system includes four stages, specifically a preprocessing stage based on a motion state model, a scale error correction stage based on laser tracking data and high-confidence motion capture data, a representative point output stage of distribution characteristics, and a data comparison calculation stage.
4. The method for measuring the spatial accuracy of an optical motion capture system according to claim 3, characterized in that: In the preprocessing stage, the measurement sequences collected by the optical motion capture system and the laser tracker are accurately matched according to the recorded timestamps, and finally the data collected by the laser tracker is used as the reference system to measure the error through the root mean square of all samples.
5. The method for measuring the spatial accuracy of an optical motion capture system according to claim 2, characterized in that: The timestamp information of the point cloud data can be used to divide the collected point cloud data into two stages based on the time period: motion start and stop, and motion stability. The data distribution of continuous time points can be simulated by Gaussian process, and the time dependency can be simulated by Gaussian function. The form of Gaussian function is as follows: Where t is time, which indicates the timestamp of the point cloud data; μ is the mean of the Gaussian function, which indicates the central time point of the stable phase of the motion; σ is the standard deviation of the Gaussian function, which controls the width of the Gaussian function; The distribution of the Gaussian function on the time axis can be controlled by adjusting the values of μ and σ; μ is set to the midpoint of the entire time period, and σ is adjusted according to the specific situation.
6. The method for measuring the spatial accuracy of an optical motion capture system according to claim 1, characterized in that: The stability of the data measured by the optical motion capture system during each pause can be expressed as the variance distribution of each point in each local point cloud on the three coordinates of the optical motion capture system measurement coordinate system; the obtained variance data reflects the degree of data dispersion of each point on the x, y and z axes. The smaller the dispersion, the more stable the measurement system. The formula is as follows: in: is the variance of the X-axis point coordinates in the nth local point cloud, x ni is the X-axis coordinate value of the i-th point in the n-th local point cloud, It is the average value of the X-axis coordinate value of the point coordinates in the nth local point cloud.
7. The method for measuring the spatial accuracy of an optical motion capture system according to claim 2, characterized in that: Each group of pause points collected by the optical motion capture system is filtered and output with representative points to obtain k representative points. (k=84), the point set obtained by laser tracking data collection is After the two point sets are processed through registration, the root mean square error (RMSE) of the Euclidean distance between the aligned points is used to measure the acquisition accuracy of the optical motion capture system; The distance error DE of the i-th observation point is: Overall root mean square error RMSE: Where: k is the number of pause points, is the measured coordinate of the optical motion capture system at the ith stop point, is the measured coordinate of the laser tracking point at the i-th stop point.
8. The method for measuring the spatial accuracy of an optical motion capture system according to claim 1, characterized in that: The point set collected when the robot arm stops is clustered by distance clustering, and the marked points are extracted into different subsets, and the center of each cluster is used as the representative point of the cluster; the mean of all cluster centers is weightedly calculated according to the number of data points in the cluster; For each cluster Ci, the mean of all points is calculated as the cluster center: Where p is a point in cluster Ci, |C i. size| is the number of data points in cluster Ci.
9. The method for measuring the spatial accuracy of an optical motion capture system according to claim 8, characterized in that: By using the cluster centers and the number of data points in each cluster, the weighted mean of all cluster centers is calculated using the number of data points as the weight: Where i is the index of the cluster, C i size is the cluster C i The number of data points in C i Center is cluster C i center.