Welding robot hand-eye calibration method based on tetrahedron model

Abstract

The invention discloses a welding robot hand-eye calibration method based on a tetrahedron model, which comprises the following steps: S1, preparing the tetrahedron model, carrying out laser scanning on a tetrahedron on a welding robot, and obtaining point cloud data of the tetrahedron through a nine-point method; s2, performing coarse registration according to the standard tetrahedron model point cloud data, and calculating to obtain T1; s3, according to a coarse registration result, performing fine registration calculation to obtain T2; and S4, performing combined calculation on the coarse registration result and the fine registration result to obtain a final pose matrix T3. According to the method, through the artificially customized tetrahedron with a specific shape and the optimization step of combining the coarse registration method and the fine registration method, the initial value is given by using the coarse registration method, then the initial value is used as input, the calculation of the rotation matrix is completed by using the fine registration method, complex plane extraction is not needed, the artificial participation is less, and the calculation efficiency is high. And the applicability is high, errors caused by a plane extraction step are avoided, high measurement precision is ensured, and a calibration result is more reliable.

Description

Technical Field

[0001] The invention relates to the technical field of vision calibration of welding robots, and in particular to a hand-eye calibration method of welding robots based on a tetrahedron model. Background Art

[0002] With the development of technologies such as 3D vision and laser radar, welding robots are being used more and more widely. The accuracy of the hand-eye relationship during the hand-eye collaborative operation of the welding robot is one of the key factors affecting the operating efficiency and accuracy of the welding robot. For the calibration of the hand-eye relationship, the current mainstream hand-eye calibration methods are:

[0003] 1> Use a tetrahedron with a specific structure and collect the 3D point cloud of the tetrahedron. Extract the three sides of the tetrahedron respectively and calculate the intersection points (vertices) of the three sides. Repeat the point cloud collection by changing the position of the robot arm and complete the calibration using the five-point method or the nine-point method. However, this method is affected by the accuracy of the extracted three sides when calculating the intersection points of the three faces, and the stability of the intersection calculation is poor.

[0004] 2> Similar to the calibration plate in camera calibration, a reflective sheet with a specific reflection intensity is artificially made, and the three-dimensional coordinates of the cross points of the reflective sheet are manually extracted, and the calibration is completed using the five-point or nine-point method. However, this algorithm involves a lot of human involvement, and its accuracy cannot be guaranteed.

[0005] 3> Using the model matching method, for a specific model (tetrahedron, sphere), etc., manually collect the point cloud data of a standard model and the point cloud data of a specific position (highest point, sphere center), etc., and then calculate the rotation matrix according to the point cloud registration method. The accuracy of this method is affected by the registration algorithm, and different registration methods vary greatly.

[0006] According to the analysis of the above hand-eye calibration method, the present invention proposes a tetrahedron-based hand-eye calibration method, which mainly adopts a two-step calibration method of coarse registration + fine registration. The coarse registration method is used to give the initial value, and then the initial value is used as input to complete the calculation of the rotation matrix using the fine registration method. This method does not require complex plane extraction and has less human involvement. After testing, it has stable accuracy and high applicability. Summary of the invention

[0007] In order to solve the technical problem that the calibration accuracy of the hand-eye calibration method in the prior art is affected by errors and is not accurate enough, the present invention provides a welding robot hand-eye calibration method based on a tetrahedron model.

[0008] To achieve the above object, the present invention provides the following technical solution: a welding robot hand-eye calibration method based on a tetrahedron model, comprising the following steps:

[0009] S1: Prepare a tetrahedron model, set up a laser scan of the tetrahedron on the welding robot, and obtain tetrahedron point cloud data using the nine-point method;

[0010] S2: Perform rough registration based on the standard tetrahedron model point cloud data and calculate T 1 ;

[0011] S3: Based on the rough registration result, perform fine registration calculation to obtain T 2 ;

[0012] S4: Combine the results of coarse and fine registration to obtain the final pose matrix T 3 .

[0013] Preferably, in step S1, a tetrahedron with a side length of 10 cm / 15 cm is first set as required, and a laser scanner is set on the movable robot arm of the welding robot, different position change modes of the movable robot arm and the laser scanner are planned, and data in nine different posture states are scanned and collected.

[0014] Preferably, in step S1, during the scanning and acquisition process, not only the point cloud data of the tetrahedron but also the data of the tool center point (TCP) of the welding gun are acquired.

[0015] Preferably, in step S1, a point cloud data Pts of a model is collected by careful scanning. (x,y,z) , and artificially set a highest point coordinate

[0016] Preferably, in step S2, based on the point cloud data collected by the nine-point method, the data features of each point in the sampled point cloud are preliminarily described by the FPFH feature method, and the relative position relationship between the point clouds is preliminarily estimated.

[0017] Preferably, in step S2, by combining the sampling consistency algorithm with the greedy algorithm, the point cloud is searched for points with high similarity to the FPFH feature of the sampling point to form a point set, and finally the rotation matrix is ​​calculated and the position of the point cloud to be registered is transformed according to the rotation and translation matrices to complete the rough registration, and a preliminary rotation and translation matrix T is obtained. 1 .

[0018] Preferably, in step S3, based on the rough registration, the GICP method is used to perform further precise adjustments on the initially aligned point cloud.

[0019] Preferably, in step S3, a more accurate rotation and translation matrix T is obtained by minimizing the distance between point clouds to optimize the target. 2 .

[0020] Preferably, in step S4, the rotation and translation matrix T obtained by the rough registration is 1The rotation and translation matrix T obtained by precise alignment 2 Combine and calculate the pose matrix T 3 =T 1 *T 2 .

[0021] Compared with the prior art, the present invention has the following beneficial effects:

[0022] The present invention adopts an optimization step combining a coarse registration method with a fine registration method by artificially customizing a tetrahedron of a specific shape. The coarse registration method is first used to give an initial value, and then the initial value is used as an input to complete the calculation of the rotation matrix by the fine registration method. There is no need to perform complex plane extraction and less human participation. The high applicability avoids the error caused by the plane extraction step, ensures high measurement accuracy, and makes the calibration result more reliable. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 A schematic diagram of the welding robot calibration site provided by the present invention;

[0024] Figure 2 It is a schematic diagram of the overall calibration process of the present invention;

[0025] Figure 3 It is a schematic diagram of the rough registration process of the present invention;

[0026] Figure 4 It is a schematic diagram of the precise registration process of the present invention.

[0027] In the figure: 1. Tetrahedron; 2. Moving robot arm; 3. Laser scanner; 4. Welding gun. DETAILED DESCRIPTION

[0028] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are 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 creative work are within the scope of protection of the present invention.

[0029] See also Figure 1 - Figure 4 The present invention provides an embodiment: a welding robot hand-eye calibration method based on a tetrahedron model, comprising the following steps:

[0030] S1: Prepare a tetrahedron model, set up a laser scan of the tetrahedron on the welding robot, and obtain tetrahedron point cloud data using the nine-point method;

[0031] S2: Perform rough registration based on the standard tetrahedron model point cloud data and calculate T1;

[0032] S3: Based on the rough registration result, perform fine registration calculation to obtain T2;

[0033] S4: Combine the results of coarse registration and fine registration to obtain the final pose matrix T.

[0034] Specifically, in step S1, a tetrahedron 1 with a side length of 10 cm / 15 cm is first set as required, and a laser scanner 3 is set on the movable robot arm 2 of the welding robot. The tetrahedron 1 will be used as the target object for laser scanning. Different position change modes of the movable robot arm 2 and the laser scanner 3 are planned to cover a sufficient range of the working space, and the data in nine different posture states are scanned and collected. In the scanning and collection process, not only the point cloud data of the tetrahedron 1 but also the data of the tool center point (TCP) of the welding gun 4 are obtained. At the same time, the point cloud data Pts of a model is carefully scanned and collected. (x,y,z) , and artificially set a highest point coordinate This coordinate may be used as a reference point in subsequent calculations or for data verification and other purposes.

[0035] Specifically, in step S2, based on the point cloud data collected by the nine-point method, the data features of each point in the sampled point cloud are preliminarily described by the FPFH feature method, and the relative position relationship between the point clouds is preliminarily estimated. Then, the sampling consistency algorithm is combined with the greedy algorithm to search for points in the point cloud with high similarity to the FPFH features of the sampling points to form a point set. Finally, the rotation matrix is ​​calculated and the position of the point cloud to be registered is transformed according to the rotation and translation matrices to complete the rough registration, and a preliminary rotation and translation matrix is ​​obtained. This matrix is ​​an approximate result that can roughly align point clouds in different poses.

[0036] Specifically, in step S3, based on the rough registration, the GICP method is used to further accurately adjust the preliminarily aligned point cloud, and a more accurate rotation and translation matrix is ​​obtained by minimizing the distance optimization target between the point clouds.

[0037] Specifically, in step S4, the rotation and translation matrix T obtained by rough registration is 1 The rotation and translation matrix T obtained by precise alignment 2 Combine and calculate the pose matrix T 3 =T 1 *T 2 .

[0038] Working principle: The present invention uses a tetrahedron of a specific shape that is customized artificially, and adopts an optimization step that combines coarse alignment with fine alignment methods. The coarse alignment method is first used to give an initial value, and then the initial value is used as input to complete the calculation of the rotation matrix using the fine alignment method. There is no need to perform complex plane extraction and less human participation. The high applicability avoids the error caused by the plane extraction step, avoids the error caused by the plane extraction step, ensures high measurement accuracy, and makes the calibration result more reliable.

[0039] It is obvious to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential features of the present invention. Therefore, the embodiments should be regarded as exemplary and non-limiting from any point of view, and the scope of the present invention is defined by the appended claims rather than the above description, and it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims are included in the present invention, and any reference numerals in the claims should not be regarded as limiting the claims involved.

Claims

1. A welding robot hand-eye calibration method based on a tetrahedron model, characterized in that: The following steps are involved: S1: Prepare a tetrahedron model, set up a laser scan of the tetrahedron on the welding robot, and obtain tetrahedron point cloud data using the nine-point method; S2: Perform rough registration based on the standard tetrahedron model point cloud data and calculate T 1 ; S3: Based on the rough registration result, perform fine registration calculation to obtain T 2 ; S4: Combine the results of coarse and fine registration to obtain the final pose matrix T 3 .

2. The method for hand-eye calibration of a welding robot based on a tetrahedron model according to claim 1, characterized in that: In step S1, firstly, a tetrahedron with a side length of 10 cm / 15 cm is set as required, and a laser scanner is set on the movable robot arm of the welding robot, different position change modes of the movable robot arm and the laser scanner are planned, and data in nine different posture states are scanned and collected.

3. The method for hand-eye calibration of a welding robot based on a tetrahedron model according to claim 2, characterized in that: In step S1, during the scanning and acquisition process, not only the point cloud data of the tetrahedron but also the data of the tool center point (TCP) of the welding gun should be obtained.

4. The method for hand-eye calibration of a welding robot based on a tetrahedron model according to claim 3 is characterized in that: In step S1, a point cloud data Pts of a model is collected by careful scanning (x,y,z) , and artificially set a highest point coordinate 5. The method for hand-eye calibration of a welding robot based on a tetrahedron model according to claim 1, characterized in that: In step S2, based on the point cloud data collected by the nine-point method, the data features of each point in the sampled point cloud are preliminarily described by the FPFH feature method, and the relative position relationship between the point clouds is preliminarily estimated.

6. The method for hand-eye calibration of a welding robot based on a tetrahedron model according to claim 5, characterized in that: In step S2, the sampling consistency algorithm is combined with the greedy algorithm to search for points in the point cloud with high similarity to the FPFH features of the sampling points to form a point set. Finally, the rotation matrix is ​​calculated and the position of the point cloud to be registered is transformed according to the rotation and translation matrices to complete the rough registration, and a preliminary rotation and translation matrix T is obtained. 1 .

7. The method for hand-eye calibration of a welding robot based on a tetrahedron model according to claim 1, characterized in that: In step S3, based on the rough registration, the GICP method is used to further fine-tune the preliminarily aligned point cloud.

8. The method for hand-eye calibration of a welding robot based on a tetrahedron model according to claim 7, characterized in that: In step S3, the target is optimized by minimizing the distance between point clouds to obtain a more accurate rotation and translation matrix T 2 .

9. The method for hand-eye calibration of a welding robot based on a tetrahedron model according to claim 1, characterized in that: In step S4, the rotation and translation matrix T obtained by rough registration is 1 The rotation and translation matrix T obtained by precise alignment 2 Combine and calculate the pose matrix T 3 =T 1 *T 2 .

Images