Measurement system external parameter calibration method based on point cloud data

Abstract

The invention discloses a measuring system external parameter calibration method based on point cloud data, and belongs to the technical field of three-dimensional vision measurement and point cloud processing. In order to solve the problems that an existing calibration technology needs to calibrate the fixed pose of an object, operation is complex and precision is limited, upper and lower surface point cloud data collected for multiple times under any pose of a standard cube block are used, a unit normal vector is obtained through plane fitting, and surface parallelism and thickness constancy are used as geometric constraints, so that the calibration precision is improved. Firstly, a pitch angle and a roll angle are optimized and solved, then a point cloud is projected into an image and a contour direction is extracted to calculate a yaw angle, and finally translation parameters are jointly optimized, so that calibration of external rotation and translation parameters of a measurement system is completed in a full-automatic and high-precision manner. The method gets rid of the limitation of a fixed pose, improves the operation flexibility and scene adaptability, and can achieve the effects of high robustness, high automation degree and good universality.

Description

Technical Field

[0001] This invention relates to the field of three-dimensional vision measurement and point cloud processing technology, and more specifically to a method for calibrating external parameters of a measurement system based on point cloud data. Background Technology

[0002] In applications such as industrial 3D measurement, robot vision positioning, and high-precision point cloud reconstruction, the external parameters of the measurement system (i.e., the pose of the sensor relative to the world coordinate system or the target object) directly affect the measurement accuracy. Traditional calibration methods usually rely on high-precision calibration boards (such as checkerboards or spherical targets) or require strict control of the orientation of the calibration object, which are complex to operate and have high requirements for the calibration environment.

[0003] In recent years, some studies have attempted to calibrate extrinsic parameters using standard geometric shapes (such as cubes and cylinders). However, existing methods generally suffer from the following problems: 1. The requirement that the standard block maintains a fixed posture during each acquisition limits practical application scenarios; 2. It is sensitive to point cloud noise and lacks robust plane fitting and normal vector consistency constraint mechanisms; 3. The inability to effectively decouple rotational parameters (pitch angle, roll angle, yaw angle) from translational parameters results in limited calibration accuracy; 4. The lack of effective utilization of the key physical constraint of thickness consistency makes it difficult to guarantee the geometric rationality of the calibration results.

[0004] Therefore, how to propose a method for calibrating the external parameters of a measurement system based on point cloud data, effectively eliminate errors in the external parameters of the system, and improve measurement consistency is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] In view of this, the present invention provides a method for calibrating the external parameters of a measurement system based on point cloud data. This method can automatically estimate the rotation parameters (α, β, γ) and translation parameters (α, β, γ) of a standard cube placed multiple times in arbitrary poses, using only the point cloud data from its upper and lower surfaces. t x , t y , t z This ensures that the upper and lower surfaces are parallel, have the same orientation, and the spacing is equal to the known thickness thk after calibration, thereby achieving high-precision external parameter calibration of the measurement system.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: A method for calibrating external parameters of a measurement system based on point cloud data includes: S1: Read the top and bottom surface point cloud data of the standard cube block collected multiple times, and organize them into a top surface point cloud list and a bottom surface point cloud list respectively. S2: Define the form of the plane equation, and perform plane fitting based on the point cloud data in the upper surface point cloud list and the lower surface point cloud list respectively, to obtain the plane equation coefficients of the upper surface and the lower surface. S3: Calculate the unit normal vectors of the upper and lower surfaces based on the coefficients of the plane equation; S4: Based on all unit normal vector pairs between the upper and lower surfaces, construct the rotation residual function and optimize the solution of pitch angle β and roll angle α; S5: Project the rotated point cloud of the upper and lower surfaces onto the XY plane to generate a binary image. Obtain the orientation angle of the outer rectangle through morphological processing and contour extraction, calculate the yaw angle γ, and combine them to obtain the complete rotation matrix R. S6: Based on the complete rotation matrix R and translation vector t=[ t x , t y , t z ] T A rigid body transformation model is constructed, and the translation parameters are optimized by using the thickness thk as a constraint. t x , t y , t z ; S7: Apply the optimized extrinsic parameters to the original point cloud data to complete the extrinsic parameter calibration of the measurement system.

[0007] Preferably, S4 includes: Construct the rotation matrix R that rotates the object by an angle β around the Y-axis and then by an angle α around the X-axis. YX (β,α); Combined with the rotation matrix, and taking all unit normal vector pairs of the upper and lower surfaces as input, define the rotation residual function; The pitch angle β and roll angle α that minimize the rotational residual function value are optimized using the least squares method.

[0008] Preferably, the rotational residual function is... rotate as follows: residual rotate = R YX (β,α) T normal lower +1; In the formula, normalupper The normalized unit normal vector of the upper surface. lower The normalized unit normal vector of the lower surface; Preferably, S5 includes: The rotation matrix R YX The point clouds of the upper and lower surfaces after (β,α) transformation are projected onto the XY plane to generate corresponding binary images; After performing morphological closing operations on the binary image, the image contour is extracted and the minimum bounding rectangle of the image contour is calculated. The principal direction angle of the minimum bounding rectangle is then obtained. Calculate the yaw angle γ based on the principal direction angles corresponding to the projected images of the upper and lower surfaces; By combining the yaw angle γ, pitch angle β, and roll angle α, the complete rotation matrix R is obtained.

[0009] Preferably, S6 includes: Construct a system containing the complete rotation matrix R and translation vector t=[ t x , t y , t z ] T Rigid body transformation model; For each point plower in the point cloud of the lower surface, calculate the distance from point p′ after rigid body transformation to the corresponding upper surface fitting plane upper. plane The distance between them is denoted as the first distance dist(p′, upper). plane ); The translation residual function is defined based on the first distance and thickness thk; The translation parameters that minimize the translation residual function value are obtained by optimization using the least squares method. t x , t y , t z .

[0010] Preferably, the formula for calculating point p′ of point plower in the lower surface point cloud after rigid body transformation is: p′=R plower+t.

[0011] Preferably, the translation residual function is... translate The formula is as follows: residual translate =dist(p′,upper plane )-thk.

[0012] As can be seen from the above technical solution, compared with the prior art, this invention discloses a method for calibrating the external parameters of a measurement system based on point cloud data. It allows a standard cube to be acquired multiple times in any pose, utilizing only the point cloud data of its upper and lower surfaces. By fitting a plane, calculating the normal vector, and introducing geometric constraints with constant thickness and surface parallelism, it first optimizes the pitch and roll angles in a step-by-step decoupling manner. Then, it combines projecting the point cloud into an image and extracting the contour direction to calculate the yaw angle. Finally, it jointly optimizes the translation parameters, thereby achieving high-precision, fully automatic calibration of rotation and translation parameters. This invention eliminates the dependence on a fixed pose of the calibration object, improving operational flexibility and scenario applicability. By deeply integrating geometric prior knowledge and image processing technology, a robust optimization model is constructed, effectively overcoming point cloud noise interference and parameter coupling problems. The entire process requires no manual intervention, achieving automated high-precision calibration, and providing an efficient, reliable, and universal solution for solving the system calibration problem in 3D vision measurement. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0014] Figure 1 A flowchart of the method provided by the present invention; Figure 2 Point cloud distribution diagrams of the upper and lower surfaces before calibration; Figure 3 This is a point cloud distribution map of the upper and lower surfaces after calibration; Figure 4 This is a schematic diagram of the minimum bounding rectangle of a binary image contour. Detailed Implementation

[0015] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0016] This invention discloses a method for calibrating external parameters of a measurement system based on point cloud data, such as... Figure 1 As shown, it includes the following steps: S1: Reference Figure 2The point cloud data of the upper and lower surfaces of a standard cube block collected multiple times are read and organized into an upper surface point cloud list and a lower surface point cloud list, respectively.

[0017] Read the point cloud file upper0.pcl from the upper surface of the standard cube acquired n times respectively. (n-1) The .pcl file sequentially stores variables pointcloud_upper0 to pointcloud_upper. (n-1) And organize them into a unified list pointclouds_upper; for the lower surface point cloud files lower0.pcl to lower (n-1) Perform the same processing on the .pcl file to obtain pointclouds_lower.

[0018] S2: Define the form of the plane equation, and perform plane fitting based on the point cloud data in the upper surface point cloud list and the lower surface point cloud list respectively, to obtain the plane equation coefficients of the upper surface and the lower surface.

[0019] In this embodiment, the plane equation is set to the form: ax + by + cz + d = 0.

[0020] For each upper surface point cloud pointcloud_upper i (i=0,…,n-1), based on the least squares method, the plane is fitted, and the residual is defined as r_plane=ax+by+cz+d, resulting in the plane coefficient array coefficients_plane_upper. i =[a i ,b i ,c i ,d i Similarly, coefficients_plane_lower is obtained by fitting the point cloud of the lower surface.

[0021] S3: Calculate the unit normal vectors of the upper and lower surfaces based on the coefficients of the plane equation.

[0022] Calculate the normal vector n=[a,b,c,1] from the plane coefficients. T And normalize to obtain the unit normal vector. upperi (i=0,…,n-1) and normal loweri (i=0,…,n-1).

[0023] S4: Based on all unit normal vector pairs between the upper and lower surfaces, construct the rotation residual function to optimize the solution of pitch angle β and roll angle α, including: S41. Construct the rotation matrix R of rotating by an angle β around the Y-axis and then by an angle α around the X-axis. YX(β,α); S42. Combining the rotation matrix, and taking all pairs of unit normal vectors of the upper and lower surfaces as input, define the rotation residual function `residual`. rotate as follows: residual rotate =cosθ+1= R YX (β,α) T normal lower +1; In the formula, normal upper The normalized unit normal vector of the upper surface. lower This is the normalized unit normal vector of the lower surface.

[0024] In an ideal state, the two normal vectors should be parallel and opposite to each other, so the cosine value cosθ is -1 and the residual tends to 0.

[0025] S43. Optimize the pitch angle β and roll angle α to minimize the rotational residual function value using the least squares method.

[0026] S5: Project the rotated point clouds of the upper and lower surfaces onto the XY plane to generate a binary image. Obtain the orientation angle of the bounding rectangle through morphological processing and contour extraction, calculate the yaw angle γ, and combine them to obtain the complete rotation matrix R, including: S51. The rotation matrix R YX The point clouds of the upper and lower surfaces after (β,α) transformation are projected onto the XY plane to generate corresponding binary images.

[0027] In this embodiment, the pixel value of the point cloud region is 255, and the rest are 0.

[0028] S52. Perform morphological closing operations on the binary image to fill holes and smooth boundaries, then extract the image contour and calculate the minimum bounding rectangle of the image contour (see reference). Figure 4 ), to obtain the main direction angle of the smallest bounding rectangle.

[0029] S53. Calculate the yaw angle γ based on the principal direction angles corresponding to the projected images of the upper and lower surfaces.

[0030] Yaw angle γ = γ1 - γ2; γ1 is the principal direction angle corresponding to the upper surface projection image, and γ2 is the principal direction angle corresponding to the lower surface projection image.

[0031] S54. Combining the yaw angle γ, pitch angle β, and roll angle α, we obtain the complete rotation matrix R=R Z (γ)R YX (β,α).

[0032] S6: Based on the complete rotation matrix R and translation vector t=[ t x , t y , t z ] T A rigid body transformation model is constructed, and the translation parameters are optimized by using the thickness thk as a constraint. t x , t y , t z ,include: S61. Construct a system containing a complete rotation matrix R and a translation vector t=[ t x , t y , t z ] T Rigid body transformation model.

[0033] S62. For each point plower in the lower surface point cloud, calculate the point p′ after rigid body transformation to the corresponding upper surface fitting plane upper. plane The distance between them is denoted as the first distance dist(p′, upper). plane ).

[0034] The formula for calculating point p′ after rigid body transformation is as follows: p′=R plower+t.

[0035] S63. Based on the first distance and thickness thk, the formula for the translation residual function is defined as follows: residual translate =dist(p′,upper plane )-thk.

[0036] S64. Optimize the translation parameters that minimize the translation residual function value using the least squares method. t x , t y , t z .

[0037] S7: Optimize the external parameters (α,β,γ,t) x ,t y ,t z This method is applied to raw point cloud data to achieve a standardized representation with aligned upper and lower surface poses and a constant spacing of thk, thus completing the calibration of external parameters for the measurement system. The results are as follows: Figure 3 As shown.

[0038] The method proposed in this embodiment is used for industrial 3D scanner calibration. In a high-precision 3D scanning system, a standard aluminum cube with a side length of 30mm and a thickness of thk=5mm is used as the calibration block. The cube is randomly placed 10 times, and point clouds of the upper and lower surfaces are collected using a structured light camera each time, resulting in a total of 20 point cloud files.

[0039] Load the point cloud and fit the plane to obtain 10 sets of upper and lower surface normal vectors; Least squares optimization yields α = -1.2°, β = 0.8°; Image projection and bounding rectangle analysis yield γ = 3.5°; Translation parameter optimization yields t x =0.12mm,t y =-0.08mm,t z =19.97mm; After the transformation was applied, the standard deviation of the distance between the upper and lower surfaces decreased from the original ±1.5mm to ±0.03mm, and the angle between the normal vectors was less than 0.1°.

[0040] The results show that the method of the present invention can effectively eliminate system external parameter errors and significantly improve measurement consistency.

[0041] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0042] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for calibrating external parameters of a measurement system based on point cloud data, characterized in that, include: S1: Read the top and bottom surface point cloud data of the standard cube block collected multiple times, and organize them into a top surface point cloud list and a bottom surface point cloud list respectively. S2: Define the form of the plane equation, and perform plane fitting based on the point cloud data in the upper surface point cloud list and the lower surface point cloud list respectively, to obtain the plane equation coefficients of the upper surface and the lower surface. S3: Calculate the unit normal vectors of the upper and lower surfaces based on the coefficients of the plane equation; S4: Based on all unit normal vector pairs between the upper and lower surfaces, construct the rotation residual function and optimize the solution of pitch angle β and roll angle α; S5: Project the rotated point cloud of the upper and lower surfaces onto the XY plane to generate a binary image. Obtain the orientation angle of the outer rectangle through morphological processing and contour extraction, calculate the yaw angle γ, and combine them to obtain the complete rotation matrix R. S6: Based on the complete rotation matrix R and translation vector t=[ t x , t y , t z ] T A rigid body transformation model is constructed, and the translation parameters are optimized by using the thickness thk as a constraint. t x , t y , t z ; S7: Apply the optimized extrinsic parameters to the original point cloud data to complete the extrinsic parameter calibration of the measurement system.

2. The method for calibrating external parameters of a measurement system based on point cloud data according to claim 1, characterized in that, S4 include: Construct the rotation matrix R that rotates the object about the Y-axis by an angle β and then about the X-axis by an angle α. YX (β,α); Combined with the rotation matrix, and taking all unit normal vector pairs of the upper and lower surfaces as input, define the rotation residual function; The pitch angle β and roll angle α that minimize the rotational residual function value are optimized using the least squares method.

3. The method for calibrating external parameters of a measurement system based on point cloud data according to claim 2, characterized in that, The rotational residual function rotate as follows: residual rotate = R YX (b,a) T normal lower +1; In the formula, normal upper The normalized unit normal vector of the upper surface. lower This is the normalized unit normal vector of the lower surface.

4. The method for calibrating external parameters of a measurement system based on point cloud data according to claim 2, characterized in that, S5 include: The rotation matrix R YX The point clouds of the upper and lower surfaces after (β,α) transformation are projected onto the XY plane to generate corresponding binary images; After performing morphological closing operations on the binary image, the image contour is extracted and the minimum bounding rectangle of the image contour is calculated. The principal direction angle of the minimum bounding rectangle is then obtained. Calculate the yaw angle γ based on the principal direction angles corresponding to the projected images of the upper and lower surfaces; By combining the yaw angle γ, pitch angle β, and roll angle α, the complete rotation matrix R is obtained.

5. The method for calibrating external parameters of a measurement system based on point cloud data according to claim 1, characterized in that, S6 include: Construct a system containing the complete rotation matrix R and translation vector t=[ t x , t y , t z ] T Rigid body transformation model; For each point plower in the point cloud of the lower surface, calculate the distance from point p′ after rigid body transformation to the corresponding upper surface fitting plane upper. plane The distance between them is denoted as the first distance dist(p′, upper). plane ); The translation residual function is defined based on the first distance and thickness thk; The translation parameters that minimize the translation residual function value are obtained by optimization using the least squares method. t x , t y , t z .

6. The method for calibrating external parameters of a measurement system based on point cloud data according to claim 5, characterized in that, The formula for calculating point p′ after rigid body transformation of point plower in the lower surface point cloud is: p′=R plower+t。 7. The method for calibrating external parameters of a measurement system based on point cloud data according to claim 5, characterized in that, Translation residual function translate The formula is as follows: residual translate =dist(p′,upper plane )-thk。

Images