A precision calibration method of an optical motion capture system in a large scene
By spatially deploying common points in the optical motion capture system and using a laser tracker and the robust least squares iterative method of the Rodrigue matrix for accuracy calibration, the positioning accuracy error problem of the optical motion capture system in large scenes is solved, and the positioning accuracy of the system is improved, especially in edge and boundary areas.
Patent Information
- Application Number
- CN202310352633.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-04-04
- Publication Date
- 2026-02-10
- Estimated Expiration
- 2043-04-04
AI Technical Summary
In large scenes, the positioning accuracy of optical motion capture systems is large, especially in edge areas and scene boundaries. Furthermore, due to the distance between the camera and the target, scale errors lead to even greater positioning accuracy errors.
A common point is set up in the space of the optical motion capture system. The common point is measured using a high-precision laser tracker. The measurement coordinates of the optical motion capture system are transformed into the coordinate system of the laser tracker through the pose transformation matrix. The accuracy is calibrated by the coordinate difference. The pose transformation matrix is solved by the robust least squares iterative method of the Rodrigue matrix. The accuracy of the system is evaluated by combining the root mean square error.
It achieves accurate and reliable precision calibration of the optical motion capture system, improves the positioning accuracy of the optical capture system in large scenes, especially in edge areas and scene boundaries, and solves the calibration difficulties caused by the small number of cameras with multiple views and the long distance between the camera and the target.
Smart Images
Figure CN116630431B_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The application relates to a precision calibration method of an optical motion capture system in a large scene and belongs to the technical field of optical motion capture systems. BACKGROUND
[0002] The optical motion capture system is a high-precision positioning system composed of multiple precise and complex optical cameras, which tracks target feature points from different angles to complete real-time capture of target actions. Before use, the optical motion capture system establishes its own coordinate system through self-calibration. At present, the establishment of the coordinate system is often completed by waving a calibration rod in the common view area of the optical motion capture system, and the calibration precision can reach millimeter level. However, the self-calibration of the optical motion capture system is a relative calibration precision, and due to long-term use of the instrument, slight vibration of the scene and other reasons, the actual positioning precision and the nominal precision at the factory often differ greatly.
[0003] In addition, in the optical motion capture system laid in a large scene, the number of cameras that can be viewed at the edge area or the scene junction area is small, and the positioning precision of the edge area or the scene junction area cannot be guaranteed; and in a large scene, the distance between the camera and the target is far, and the system positioning precision error caused by the scale error is larger. SUMMARY
[0004] The application aims to provide a precision calibration method of an optical motion capture system in a large scene to solve the problem of large precision error of the optical motion capture system in a large scene.
[0005] The application provides a precision calibration method of an optical motion capture system in a large scene, which comprises the following steps:
[0006] 1) Selecting a common point and laying it out;
[0007] 2) Measuring the common point by using a laser tracker and an optical motion capture system, and determining the pose transformation matrix of the optical motion capture system and the laser tracker by using the measurement data of the common point;
[0008] 3) Converting the common point coordinates measured by the optical motion capture system to the coordinate system of the laser tracker by using the pose transformation matrix, and differencing the converted coordinates and the measurement results of the laser tracker, and calibrating the precision of the optical motion capture system according to the difference results.
[0009] The present application can accurately and reliably realize precision calibration of the optical motion capture system through the above process.
[0010] Further, if the optical motion capture system is distributed in multiple regions in space and the number of camera regions that can be observed in common between the regions is less than a set threshold in the step 1), the common points are arranged in each region.
[0011] Further, when the common points are arranged in each region, the optical motion capture system is first calibrated in terms of regional precision using the common points in each region, and then the optical motion capture system is calibrated in terms of overall precision using the common points in the entire region.
[0012] The present application improves the accuracy of precision calibration of the optical capture system in a large scene by separately calibrating each region to determine the calibration precision of the optical capture system in each region and then calibrating the optical motion capture system in the entire region.
[0013] Further, when the common points are arranged, the common points are uniformly arranged in the corresponding space and distributed in the entire region that can be observed by the optical motion capture system.
[0014] When the common points are arranged, the present application uniformly arranges the common points in the entire region that can be observed by the optical motion capture system to ensure the precision of subsequent calculations.
[0015] Further, when the common points are measured by the laser tracker and the optical motion capture system in the step 2), the same size and same ball center position of the tracker target ball and the retro-reflective marker target ball are selected as tools.
[0016] The present application selects the same size and same ball center position of the retro-reflective marker target ball and the tracker target ball as tools to solve the problem of difficulty in measuring common points by different systems and improve the precision of common point measurement.
[0017] Further, when the common points are measured, the tracker target ball is placed at the common point position, the coordinates of the common point in the tracker coordinate system are measured by the laser tracker, then the base position is fixed, the retro-reflective marker target ball is used to replace the tracker target ball, and the coordinates of the retro-reflective marker target ball are measured by the optical motion capture system.
[0018] Further, the pose transformation matrix in the step 2) is solved by using the robust least square iteration method based on the Rodrigues matrix.
[0019] The present application solves the common point conversion matrix by using the robust least square iteration method based on the Rodrigues matrix, which has less parameters to be estimated, high calculation accuracy and avoids the influence of gross errors on the result through the robust estimation.
[0020] Further, the precision is represented by the root mean square error, and the calculation formula is:
[0021]
[0022] Wherein P e represents the difference between the coordinates of the common point converted to the coordinate system of the laser tracker and the measurement result of the laser tracker, n represents the number of common points, e represents the root mean square error, i.e. the precision of the optical motion capture system.
[0023] The present application represents the precision of the optical motion capture system by the root mean square error of the difference between the coordinates of the common point converted to the coordinate system of the laser tracker and the measurement result of the laser tracker, which can accurately and quickly determine the positioning precision of the optical motion capture system. BRIEF DESCRIPTION OF DRAWINGS
[0024] Figure 1 is the flow chart of the precision calibration method of the optical motion capture system in a large scene of the present application;
[0025] Figure 2 is the measurement principle diagram of the laser tracker;
[0026] Figure 3 is the camera distribution and common point layout schematic diagram of the optical motion capture system in a large scene;
[0027] Figure 4 is the common point layout scheme schematic diagram in the present application;
[0028] Figure 5 is the common point measurement schematic diagram of the present application. DETAILED DESCRIPTION
[0029] The specific embodiments of the present application will be further described in combination with the drawings.
[0030] This invention establishes common points in the space surrounding the optical motion capture system, measures these common points using a high-precision laser tracker, and determines the pose transformation matrix between the optical motion capture system and the laser tracker based on the coordinates of these common points. The pose transformation matrix is then used to convert the coordinates of the common points measured by the optical motion capture system to the coordinate system of the laser tracker. Finally, the accuracy of the optical motion capture system is calibrated based on the difference between the converted coordinates and the measurement results from the laser tracker. The implementation process of this method is as follows: Figure 1 As shown below, a detailed explanation will follow.
[0031] 1. Select common points and set them up.
[0032] Several common points are planned within the field of view of the motion capture system, such as Figure 4 As shown. The common points should be evenly distributed spatially, covering the entire area observable by the optical motion capture system. The common points should have a certain distribution in the X, Y, and Z directions. For optical motion capture systems deployed in large scenes, if the system is distributed across multiple spatial regions and the number of camera regions that can be viewed simultaneously across these regions is small, then regional precision calibration is required. Therefore, the common points should also be deployed regionally, such as... Figure 3 As shown, in this embodiment, the optical motion capture system is distributed in three areas, and the common points are also arranged according to the three areas.
[0033] 2. Measure the common point using a laser tracker and an optical motion capture system respectively.
[0034] Currently, high-precision measuring instruments capable of reaching sub-millimeter levels include laser trackers, coordinate measuring machines (CMMs), and micrometers. However, CMMs and micrometers employ contact measurement methods, resulting in limited measurement ranges, making them unsuitable for calibrating the accuracy of optical motion capture systems in large-scale environments. Laser trackers are a type of high-precision measuring instrument. Currently, advanced laser trackers on the market, such as the Leica AT901, offer a ranging accuracy of 15µm + 6ppm, a maximum measuring range of 80 meters, and automatic aiming functionality. They can measure the positioning accuracy of motion capture systems. The measurement principle of laser trackers is as follows: Figure 2 As shown.
[0035] Since the optical motion capture system measures the center of the marker sphere, while the tracker measures the center of the target sphere, the points measured by the two systems are not concentric, i.e., they are not common points. Therefore, selecting a suitable measuring device to unify the common point of the two systems is a challenge in measurement. To solve the problem of inconsistent measurement points, a Leica series hemispherical photogrammetric reflective workpiece of the same size is used to replace the marker sphere of the optical motion capture system. Because the photogrammetric reflective workpiece and the tracker's target sphere have the same radius, and the center of the photogrammetric reflective workpiece coincides with the center of the tracker's target sphere, the center of the target sphere measured by the tracker and the center of the photogrammetric reflective workpiece measured by the optical motion capture system are the same measurement point, achieving the goal of measuring the common point of the two systems.
[0036] During measurement, such as Figure 5 As shown, the target ball of the tracker is placed at the common point A. The tracker records the position of A in the tracker coordinate system. Then, a photogrammetric marker ball is used to replace the target ball of the tracker. The base position remains unchanged, and the optical motion capture system obtains the coordinates of the photogrammetric marker ball. Then, the target base is placed at point B. The tracker and the optical motion capture system obtain the position coordinates of point B in sequence. The measurement is repeated until the position coordinates of the common points are measured in sequence.
[0037] 3. Determine the pose transformation matrix of the optical motion capture system and the laser tracker using the measurement data of the common points.
[0038] The Rodrigue matrix, consisting of three independent elements, offers advantages such as simple computation, high accuracy, and avoids the problem of determining the quadrant of the rotation angle. Therefore, an iterative method based on the Rodrigue matrix is used to solve for the coordinate system transformation matrix. The definition of the Rodrigue matrix is as follows:
[0039] (1) Introducing the matrix Where a, b, and c are independent.
[0040] (2) Define I as a third-order identity matrix, then R = (I + S)(IS) -1 It is an orthogonal matrix. Expanding R, we get:
[0041]
[0042] Suppose that the coordinates of a certain point in the tracker coordinate system and the optical motion capture system coordinate system are (x, y, y) and (x, y, y) respectively. i ,y i ,z i ) T and(u i ,v i ,w i ) T We can obtain:
[0043]
[0044] Where R is the rotation matrix, [T] x T y T z ] T It is a translation matrix.
[0045] The difference between the common points in the two coordinate systems is obtained by subtracting:
[0046]
[0047] The above equation, when simplified by left-multiplying by (IS), yields:
[0048]
[0049] Let k ij =k i -k j The above equation simplifies to:
[0050]
[0051] Substituting I and S, and simplifying, we get:
[0052]
[0053] In the form Ax = B, since A is a singular matrix, at least two systems of equations are needed to solve for a, b, and c. Therefore, when there are n common points, a 3(n-1) × 3 dimensional system of equations can be obtained as follows:
[0054]
[0055] The initial values of the unknowns a0, b0, c0 are obtained using the least squares principle. The initial values of the rotation matrix R0 and the translation matrix T are then obtained according to formulas (1) and (2), respectively. x0 ,T y0 ,T z0 ) T .
[0056] From formulas (1) and (2), we get:
[0057]
[0058] Linearization yields:
[0059]
[0060] Therefore, the error equation is:
[0061] v = Mx - N (10)
[0062] in
[0063]
[0064]
[0065]
[0066] For calculations of the form MX=N, the unknown parameters are typically calculated using least squares iteration. However, due to the large number of common points collected, gross errors are inevitable. To avoid the impact of gross errors on parameter estimation, robust estimation based on IGG3 is introduced here for least squares estimation.
[0067] From formula (10):
[0068]
[0069] in, This is an equivalent weight. Since each observation of a common point is independent, therefore It is a diagonal matrix. Each diagonal element in the array satisfies:
[0070]
[0071] P i The weight factor coefficient for each observation is set to 1. According to the definition in IGG3, w i The form is:
[0072]
[0073] Where k0 is typically taken as 1.0 to 1.5, k1 is typically taken as 3.0 to 4.5, and v is the residual v i Ratio to standardized residuals The equation for calculating the standardized residual is:
[0074]
[0075] in, The error is the unit weight. Let be the i-th element on the diagonal of the covariance matrix. The formula for calculating the unit weight variance is:
[0076]
[0077] Where V represents the error term, P represents the weight matrix, m represents the number of rows in the system of equations 3(n-1), and t represents the number of parameters to be estimated, which is taken as 6.
[0078] In summary, the calculation process is as follows: First, the initial values of the unknowns a0, b0, c0 are obtained by solving the least squares principle; second, the initial values of the rotation matrix R0 and the translation matrix (T) are calculated. x0 ,T y0 ,T z0 ) T Then, combining the initial value R0 and (T) x0 ,T y0 ,T z0 ) T Linearization establishes unknowns a, b, c, and T. x ,T y ,T z The error equation is solved by robust least squares iteration to obtain the unknowns a, b, c and T. x ,T y ,T z The rotation matrix R and translation matrix T are obtained by solving.
[0079] 4. Use the obtained pose transformation matrix for accuracy calibration.
[0080] Let the coordinates of the common point in the optical motion capture system and the laser tracker system be P and P, respectively. V and P l The solved pose transformation matrix of the optical motion capture system and the laser tracker is: and ( For rotation matrix, (The translation matrix is used to transform the coordinates of the common point measured by the optical motion capture system to the coordinate system of the laser tracker using the obtained pose transformation matrix, thus obtaining P.) l Then, the transformed common point coordinates are compared with the coordinates P actually measured by the laser tracker. L The specific formula used for calculating the difference is as follows:
[0081]
[0082] Using the obtained P e To determine the positioning accuracy of the optical motion capture system, this embodiment calculates the root mean square error (RMSE) of the system's points. The obtained RMSE is then used to evaluate the positioning accuracy of the optical motion capture system. The formula for calculating the RMSE of the optical motion capture system is as follows:
[0083]
[0084] Where P e This represents the difference between the coordinates of the common point transformed into the coordinate system of the laser tracker and the measurement result of the laser tracker. n represents the number of common points, and e represents the root mean square error, which is the accuracy of the optical motion capture system.
[0085] It is worth noting that for optical motion capture systems deployed in large-scale scenes, if the optical capture system is distributed across multiple spatial regions and the number of camera areas that can be viewed simultaneously across these regions is relatively small, the accuracy calibration of the optical capture system in each region should be performed separately before performing an overall accuracy calibration of the entire optical motion capture system for the entire region. For example... Figure 3 As shown, in this embodiment, the cameras of the optical motion capture system in a large scene are distributed in three regions. The cameras in each region have good common-view conditions, and a small number of cameras in different regions share a common-view relationship. For the accuracy calibration of the optical motion capture system in a large space, the transformation matrix between the tracker coordinate system and the coordinate system of the optical motion capture system in each region should be solved first, and the positioning accuracy of each region of the optical motion capture system should be evaluated separately. Then, the common points of each region should be grouped together to evaluate the overall positioning accuracy of the optical motion capture system. The purpose of regional calibration is that the system accuracy of regional and overall regions often differs greatly in large scenes. In each regional region, due to the large number of cameras sharing a common-view relationship and the close distance between the cameras and the target, the system accuracy in each regional region is often higher. In the overall region, due to the small number of cameras sharing a common-view relationship and the far distance between the cameras and the target, the system accuracy in the overall region may be significantly reduced. Therefore, by performing overall region calibration and regional calibration in large scenes, the positioning accuracy of the motion capture system can be evaluated more objectively and comprehensively, achieving a more accurate and objective accuracy calibration of the optical motion capture system.
[0086] Specifically, the process is as follows: Based on the common points of Region 1, the coordinate transformation matrix between the tracker coordinate system and the optical motion capture system coordinate system of Region 1 is obtained. The transformation error of the common points in Region 1 is then calculated using the coordinate transformation matrix to obtain the positioning accuracy of the optical motion capture system in Region 1. Similarly, the positioning accuracy of the optical motion capture system in Regions 2 and 3 is calculated based on the common points of Regions 2 and 3, respectively. Finally, the common points of the three regions are combined, and the coordinate transformation matrix between the tracker coordinate system and the overall optical motion capture system coordinate system is calculated. The positions of all common points in the optical motion capture system coordinate system are then converted to the laser tracker coordinate system coordinate system and subtracted from the positions of the common points measured by the tracker to calculate the positioning accuracy of the overall optical motion capture system.
[0087] The calibration method proposed in this invention is applicable to the accuracy calibration of optical motion capture systems in large and complex scenes. It can solve the problems of small shared viewing areas of cameras in different regions of the optical motion capture system, large distances between cameras and targets, and a small number of shared viewing cameras, which lead to calibration difficulties. By calibrating each region separately to determine the calibration accuracy of the optical capture system in each region, and then performing overall accuracy calibration of the optical motion capture system in the entire region, the accuracy calibration of optical capture systems in large scenes can be achieved accurately and reliably.
Claims
1. A method for calibrating the accuracy of an optical motion capture system in a large-scale scene, characterized in that, The calibration method includes the following steps: 1) Select and deploy common points. If the optical motion capture system is distributed in multiple areas of space and the number of camera areas that can be viewed by each area is less than the set threshold, then deploy common points in each area. When each area has corresponding common points, first use the common points of each area to perform regional accuracy calibration of the optical motion capture system, and then use the common points of the entire area to perform overall accuracy calibration of the optical motion capture system. 2) Measure the common points using both the laser tracker and the optical motion capture system, and use the measurement data of the common points to determine the pose transformation matrix of the optical motion capture system and the laser tracker; When determining the pose transformation matrix, the initial value of the pose transformation matrix is first calculated according to the least squares method. Then, the error equation of the initial value of the pose transformation matrix is established by linearization. The error equation is solved by the IGG3 robust estimation method and least squares iteration to obtain the pose transformation matrix. 3) Use the pose transformation matrix to transform the coordinates of the common point measured by the optical motion capture system to the coordinate system of the laser tracker, and calculate the difference between the transformed coordinates and the measurement results of the laser tracker. Based on the difference result, calibrate the accuracy of the optical motion capture system.
2. The accuracy calibration method for a large-scene optical motion capture system according to claim 1, characterized in that, The pose transformation matrix is obtained by an iterative method based on the Rodrigue matrix.
3. The accuracy calibration method for a large-scene optical motion capture system according to claim 1, characterized in that, The pose transformation matrix includes a rotation matrix and a translation matrix.
4. The accuracy calibration method for a large-scene optical motion capture system according to claim 1, characterized in that, When setting up common points, they are evenly distributed in the corresponding space and within the entire area that the optical motion capture system can observe.
5. The accuracy calibration method for a large-scene optical motion capture system according to claim 1, characterized in that, In step 2), when measuring the common point using a laser tracker and an optical motion capture system, tracker target balls and reflective target balls of the same size and center position are selected as tools.
6. The accuracy calibration method for a large-scene optical motion capture system according to claim 5, characterized in that, When measuring the common point, the tracker target ball is placed at the common point location, and the coordinates of the common point in the tracker coordinate system are measured using a laser tracker. Then, the base position remains unchanged, and a reflective marker target ball is used instead of the tracker target ball. The optical motion capture system measures the coordinates of the reflective marker target ball.
7. The accuracy calibration method for a large-scene optical motion capture system according to claim 1, characterized in that, The accuracy mentioned is characterized by the root mean square error, and the calculation formula is as follows: ； Where P e This represents the difference between the coordinates of the common point transformed into the coordinate system of the laser tracker and the measurement result of the laser tracker. n represents the number of common points, and e represents the root mean square error, which is the accuracy of the optical motion capture system.