An online calibration method for spatial kinematic chain parameters of a robot 3D vision measurement system

Abstract

A method for online calibration of space kinematic chain parameters of a robot three-dimensional visual measurement system belongs to the field of intelligent robot technology, and includes S1: constructing a space kinematic chain model based on a virtual end; S2: using a PnP algorithm to calculate the position and direction of the robot base coordinate system relative to the measured target coordinate system, and obtaining the kinematic chain parameters under the unified model of the space kinematic chain based on the virtual end; S3: based on the unified model of the space kinematic chain based on the virtual end constructed in step S1 and the kinematic chain parameters from the measured target coordinate system to the robot base coordinate system solved in step S2, a new optimization function is constructed, and a nonlinear iterative optimization method is used to optimize the parameters of the unified model of the space kinematic chain based on the virtual end. The method of the invention is based on the traditional hand-eye calibration kinematic model of AX=XB, and uses a new unified model of the space kinematic chain based on the virtual end to realize the online calibration of the space kinematic chain parameters of the robot three-dimensional visual measurement system, and the model can be applied in practical engineering.

Description

Technical Field

[0001] The invention belongs to the technical field of intelligent robots and relates to a method for online calibration of spatial motion chain parameters (hand-eye parameters+robot kinematic parameters) of a robot three-dimensional visual measurement system under actual operating conditions. Background Art

[0002] Industrial robots are key supporting equipment for intelligent manufacturing. Robot 3D vision measurement technology is an important sensing method for intelligent manufacturing factories. It can automatically, accurately and quickly obtain key 3D dimensions in the process of parts manufacturing, and is of great significance for defect detection, geometric positioning, and autonomous execution in the manufacturing process. For example, automated measurement of large-scale parts and skins in aerospace (for rapid quality inspection and assembly planning), online measurement in the surface grinding and polishing of high-speed rail body-in-white (for real-time planning of processing trajectories), and robot-assisted large-plate welding in shipbuilding (for accurate positioning of welds and post-inspection) all require high-precision robot 3D vision measurement.

[0003] The basic structure of the robot 3D vision measurement system is that the industrial robot drives the end-mounted 3D vision measurement sensor, assisted by a turntable or linear motion mechanism, to move and scan the part to be measured, and accurately reconstruct the surface geometry of the part to be measured in 3D. In most application scenarios, the shape and size of the part to be measured are larger than the single measurement field of view of the 3D vision measurement system used. Therefore, multiple measurements from multiple angles (viewpoints) are required to fully measure the part to be measured. The process of converting 3D point cloud data measured at different angles into a unified coordinate system to build a complete geometric model is called multi-viewpoint cloud stitching, which is a core technology of the robot 3D vision measurement system.

[0004] There are currently two mainstream multi-view point cloud stitching methods: multi-view stitching based on marker points and multi-view stitching without marker points based on the robot spatial motion chain parameters. In the multi-view stitching method based on marker points, it is necessary to paste artificial marker points on the surface of the part to be measured to achieve auxiliary stitching, and the marker points need to be removed after the measurement is completed. Sandro Barone et al. [Barone S, Paoli A, Razionale A V. Three-dimensional point cloud alignment detecting fiducial markers by structured light stereo imaging [J]. Machine Vision & Applications, 2012, 23 (2): 217-229.] systematically discussed the multi-view stitching method based on marker points. Although this method can guarantee the accuracy of stitching to a certain extent, the pasting process of artificial marker points greatly increases the measurement workload. Due to the certain thickness of the marker points themselves, the three-dimensional morphology of the surface of the part to be measured is distorted. In addition, some soft, wet, and hot objects to be measured are not suitable for pasting any markers on the surface. Therefore, although the multi-view stitching method based on landmark points has the advantages of clear operation process and easy application, the pasting of artificial marks also affects the measurement accuracy and limits the objects of application.

[0005] In order to realize the multi-view cloud stitching without markers of the robot 3D vision measurement system, the industry generally adopts multi-view stitching based on spatial motion chain parameters. The basic idea is to accurately calibrate the spatial motion chain parameters (hand-eye parameters + robot kinematics) of the entire robot 3D vision measurement system, and transfer the measurement data of each viewpoint to a unified world coordinate system through the accurately calibrated spatial motion chain parameters, thereby realizing multi-view cloud stitching. The general coordinate transformation flow is: sensor coordinate system → robot end tool coordinate system → robot base coordinate system. This method does not require the pasting of markers and does not rely on any external expensive tracking and positioning equipment. However, due to the long-term working conditions of the robot 3D vision measurement system, mechanical deformation, looseness, wear and other unfavorable conditions will occur, and the spatial motion chain parameters of the system will be seriously inaccurate, such as hand-eye relationship misalignment, robot kinematic parameter drift, etc. The resulting spatial motion chain parameter inaccuracy is usually at the millimeter level, which will seriously affect the accuracy of multi-view stitching of the measurement point cloud. Although, the parameters can be recalibrated under the condition of system shutdown through an external laser tracker. Xuanchen Zhang et al. [Zhang X, Song Y, Yang Y, et al. Stereo vision based autonomous robot calibration [J]. Robotics and Autonomous Systems, 2017, 93: 43-51.] systematically described the multi-view stitching method without landmarks based on the robot spatial motion chain parameters. However, the downtime of dozens of minutes will seriously affect the manufacturing rhythm and is not suitable for promotion and application in actual operating conditions.

[0006] In summary, under actual operating conditions, the spatial kinematic chain parameters of the robot 3D vision measurement system will be seriously inaccurate, resulting in reduced precision of multi-view measurement point cloud splicing. Therefore, it is necessary to develop a truly feasible online calibration method for the spatial kinematic chain parameters of the robot 3D vision measurement system to improve the precision of point cloud splicing. Summary of the invention

[0007] In response to the above defects or improvement needs of the prior art, the present invention provides an online calibration method for the spatial motion chain parameters of a robot three-dimensional vision measurement system. The method performs multi-view stitching based on the spatial motion chain parameters, constructs a new spatial motion chain model based on the virtual end and defines its cost function, and on this basis, solves and optimizes the key parameters of the spatial motion chain of the measurement system to achieve online calibration of the key parameters.

[0008] In order to achieve the above object, the technical solution adopted by the present invention is:

[0009] A method for online calibration of space kinematic chain parameters of a robot three-dimensional vision measurement system comprises the following steps:

[0010] Step S1: On the basis of the existing AX=XB traditional kinematics model, a new spatial kinematic chain model based on the virtual end is constructed.

[0011] The complete space motion chain model based on the virtual end includes the robot base, the four-axis robot link 1-link 4 (the number of joints is not less than 3, the joints are joints J1-J4 in the embodiment, and the robot key joint is specifically referred to as joint J3), the three-dimensional vision measurement sensor, the measured target, the virtual end and its virtual arm (virtual arm 4', virtual arm 5'). The space motion chain model has the robot base coordinate system, the robot key joint coordinate system, the three-dimensional vision measurement sensor coordinate system, the measured target coordinate system, and the virtual end coordinate system in space.

[0012] The virtual end is connected to the robot's key joints and the three-dimensional vision measurement sensor through a virtual arm link, and the position and direction are determined by the relative geometric relationship between the Z axis of the robot's key joint coordinate system and the Z axis of the three-dimensional vision measurement sensor coordinate system, as follows:

[0013] When the relative geometric relationship between the Z axis of the robot's key joint coordinate system and the Z axis of the 3D vision measurement sensor coordinate system is parallel, the virtual end center point position is selected at any position along the Z axis of the 3D vision measurement sensor, and the virtual end coordinate system has the same direction as the 3D vision measurement sensor coordinate system;

[0014] When the relative geometric relationship between the Z axis of the robot's key joint coordinate system and the Z axis of the three-dimensional vision measurement sensor coordinate system is an intersection, the virtual end center point position is selected at the intersection of the Z axis of the three-dimensional vision measurement sensor coordinate system and the Z axis of the robot's key joint coordinate system, the Z axis of the virtual end coordinate system is along the Z axis direction of the three-dimensional vision measurement sensor coordinate system, and the X axis of the virtual end coordinate system is perpendicular to the plane formed by the two intersecting Z axes;

[0015] When the relative geometric relationship between the Z axis of the robot's key joint coordinate system and the Z axis of the three-dimensional vision measurement sensor coordinate system is inclined, the virtual end center point position is selected at a point along the Z axis of the three-dimensional vision measurement sensor coordinate system, at which the straight line perpendicular to the Z axis of the robot's key joint coordinate system intersects with the z axis of the three-dimensional vision measurement sensor coordinate system.

[0016] The position and direction of the virtual end are determined, and a new spatial motion chain model based on the virtual end is constructed.

[0017] S2: Calculate the kinematic chain parameters of the spatial kinematic chain model based on the virtual end constructed in S1 Among them, the kinematic chain parameters refer to the rotation and translation relationship from the measured target coordinate system to the robot base coordinate system. In this spatial kinematic chain model, the origin of the 3D vision measurement sensor is regarded as a virtual end, as follows:

[0018] S2.1: Calculate the positional relationship from the measured target coordinate system to the robot base coordinate system

[0019] The robot joint J1 is the only variable of the robot and rotates any n (n>3) different angles. All other joint angles remain fixed to ensure that the target is always completely within the scanning field of view of the 3D vision measurement sensor. The robot joint J1 is rotated any n different angles to obtain n different postures. Based on the images of the target captured by the 3D vision measurement sensor in n different postures, the PnP algorithm is used to calculate the positional relationship from the coordinate system of the target to the coordinate system of the robot base.

[0020] S2.2: Determine the orientation of the robot base coordinate system relative to the target coordinate system

[0021] The rotation angle of robot joint J1 is fixed at 0°, and robot joint J2 is rotated to any m (m>3) different angles as the only variable of the robot. All other joint angles remain fixed to ensure that the target to be measured is always completely within the scanning field of view of the 3D vision measurement sensor. Rotate robot joint J2 to any m different angles to obtain m different postures. The position of the virtual end in each posture is obtained by reverse calculation based on the target coordinate system and the target image captured by the 3D vision measurement sensor through the PnP algorithm. In this way, the direction of the robot base coordinate system relative to the target coordinate system is determined.

[0022] S2.3: Based on the position and direction of the robot base coordinate system relative to the measured target coordinate system, the kinematic chain parameters of the spatial kinematic chain model based on the virtual end are obtained.

[0023] S3: Based on the spatial kinematic chain model based on the virtual end constructed in step S1 and the kinematic chain parameters from the measured target coordinate system to the robot base coordinate system solved in step S2 As a basis, a new optimization function is constructed, and the nonlinear iterative optimization method is used to optimize the parameters of the spatial kinematic chain model based on the virtual end, as follows:

[0024] In step S2.1, the robot joint J1 rotates any n different angles to obtain n different postures, and the three-dimensional visual measurement sensor captures n different views of the target. For the robot's i-th (i=1,2……,n) posture, define is the pose transformation relationship from the 3D vision measurement sensor coordinate system to the measured target coordinate system obtained by the PnP algorithm; definition It is the position transformation relationship from the robot base coordinate system to the 3D vision measurement sensor coordinate system; define is the position transformation relationship from the robot base coordinate system to the measured target coordinate system. There is a formula:

[0025]

[0026] definition is the pose transformation relationship from the robot base coordinate system to the target coordinate system The rotational component of is the pose transformation relationship from the robot base coordinate system to the target coordinate system The translation component of R z The kinematic chain parameters from the measured target coordinate system to the robot base coordinate system The rotational component of z The kinematic chain parameters from the measured target coordinate system to the robot base coordinate system The translation component. Using Rodrigues's rotation formula Rodrigues (R), for n different postures, the position error PError and orientation error OError of the spatial kinematic chain model based on the virtual end are:

[0027]

[0028] Then the optimization function of the spatial kinematic chain model parameters based on the virtual end is:

[0029]

[0030] Among them, a l , d l , α l ,θ l They correspond to the link length, link offset, link twist and joint angle of the lth link in the space motion chain model based on the virtual end, respectively, where l = 1, 2, 3, 4, 5: 1-3 represent link 1, link 2, link 3 respectively; 4, 5 represent virtual arm 4', virtual arm 5' respectively. Based on the optimization function solved in step S3, after optimizing the parameters of the space motion chain model based on the virtual end using the nonlinear iterative optimization method, the three-dimensional vision measurement sensor rescans the measured object to obtain the optimized data, and uses the optimized data as the initial value for re-optimization.

[0031] The maximum expected deviation threshold of the kinematic parameters of the spatial kinematic chain model of the virtual end is set to (e a , e d , e α , eθ ), and the optimization process is repeated using nonlinear iterative optimization until |a l | <e a ,|d l | <e d ,|α l | <e α ,|θ l | <e θ When , the parameter optimization of the space kinematic chain model based on the virtual terminal is completed. If it is still not optimized to below the threshold, it is necessary to return to step S2 to recalculate the kinematic chain parameters under the space kinematic chain model based on the virtual terminal and optimize them.

[0032] The present invention constructs a space motion chain model based on a virtual end, defines a new optimization function, uses a nonlinear iterative optimization method to optimize the parameters of the space motion chain model based on the virtual end, and finally realizes an algorithm that is not affected by the parameter errors of the robot motion chain.

[0033] In general, the above technical solutions conceived by the present invention can achieve the following beneficial effects compared with the prior art:

[0034] (1) Compared with the multi-view stitching method without marker points based on the robot spatial kinematic chain parameters, this method requires dozens of minutes of downtime operation to recalibrate the robot parameters through an external laser tracker under the condition of system shutdown, which will seriously affect the manufacturing rhythm and is not suitable for promotion and application in actual operating conditions. The initial calibration process of the spatial kinematic chain model newly constructed by the present invention based on the virtual end hardly involves the posture capture process and only takes 82 seconds. In the continuous operation condition, the subsequent recalibration and downtime operation time are less than 82 seconds, which greatly improves the work efficiency.

[0035] (2) Regarding the accuracy, in the multi-view stitching method without landmarks based on the robot's spatial kinematic chain parameters, the errors in the intrinsic parameters of the 3D vision measurement sensor and the robot's spatial kinematic chain parameters will lead to inaccurate hand-eye transformation, resulting in a high error in 3D data stitching. A 1% error in the spatial kinematic chain parameters will cause a 1.1638mm data stitching error. Experimental results show that the online calibration method for the spatial kinematic chain parameters of the robot's 3D vision measurement system effectively maintains the average stitching error at <0.1mm under a 1% error in the spatial kinematic chain parameters, and the accuracy is less affected by factors such as temperature changes. The ability to resist the interference of spatial kinematic chain parameter errors and the characteristics of rapid recalibration make the online calibration method for the spatial kinematic chain parameters of the robot's 3D vision measurement system proposed in this paper a reality in practical applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1A schematic diagram of a space kinematic chain model based on a virtual terminal provided in an embodiment of the present invention;

[0037] Figure 2 is a schematic diagram of obtaining a reference coordinate system provided in an embodiment of the present invention, in which Figure 2(a) uses a circular trajectory formed by a virtual end in three-dimensional space to determine the position of a robot base coordinate system relative to a measured target coordinate system; Figure 2(b) uses a linear trajectory formed by a virtual end in three-dimensional space to determine the direction of a robot base coordinate system relative to a measured target coordinate system.

[0038] Figure 3 The present invention provides an overall flow chart of the embodiment of the present invention.

[0039] In the figure: 1 robot key joints; 2 four-axis robot link 4; 3 four-axis robot link 3; 4 four-axis robot link 2; 5 four-axis robot link 1; 6 virtual arm 4'; 7 virtual end; 8 virtual arm 5'; 9 three-dimensional vision measurement sensor; 10 robot base; 11 measured target; 12 calibration plate; ① robot base coordinate system; ② robot key joint coordinate system; ③ virtual end coordinate system; ④ three-dimensional vision measurement sensor coordinate system; ⑤ measured target coordinate system. DETAILED DESCRIPTION

[0040] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0041] Figure 1 Schematic diagram of a spatial kinematic chain model based on a virtual terminal provided in an embodiment of the present invention. Figure 1 As shown, it specifically includes: a robot base 10, a four-axis robot link 1-link 4 (5-2), joints J1-J4, the robot key joint is specifically referred to as joint J3, the base is connected to joint J1 through link 1, joint J1, joint J2, joint J3, joint J4 are connected through link 2, link 3, link 4 respectively, joint J4 is connected to a three-dimensional vision measurement sensor 9, a measured target 11, a virtual end 7 and its virtual arm (virtual arm 4'6, virtual arm 5'8), the virtual arm 4'6 connects the key joint J3 and the virtual end 7, and the virtual arm 5' connects the virtual end 7 to the three-dimensional vision measurement sensor 9.

[0042] S1: In the process of constructing the spatial kinematic chain model based on the virtual end, a virtual end is designed to be inserted between the robot's key joint 3 and the three-dimensional vision measurement sensor. The virtual end is connected to the robot's key joint and the three-dimensional vision measurement sensor through a virtual arm link. The conversion relationship from the three-dimensional vision measurement sensor coordinate system to the robot base coordinate system can be expressed as:

[0043]

[0044] in It is determined by many factors such as the length of the connecting rod p robot arm, offset, degree of distortion, joint angle, etc. They are the coordinate transformations between the robot's key joint 3 and the virtual end, and between the virtual end and the three-dimensional visual measurement sensor, which are determined by the robot's spatial motion chain parameters.

[0045] The position and direction of the virtual end are determined by the relative geometric relationship between the Z axis of the robot's key joint coordinate system and the Z axis of the three-dimensional vision measurement sensor coordinate system. When the relative geometric relationship is parallel, the virtual end center point position is selected at any position along the Z axis of the three-dimensional vision measurement sensor coordinate system, and the virtual end coordinate system and the three-dimensional vision measurement sensor coordinate system have the same direction; when the relative geometric relationship is intersecting, the virtual end center point position is selected at the intersection of the Z axis of the three-dimensional vision measurement sensor coordinate system and the Z axis of the robot's key joint coordinate system, the Z axis of the virtual end coordinate system is along the Z axis direction of the three-dimensional vision measurement sensor coordinate system, and the X axis of the virtual end coordinate system is perpendicular to the plane formed by the two intersecting Z axes; when the relative geometric relationship is inclined, the virtual end center point position is selected at a point along the Z axis of the three-dimensional vision measurement sensor coordinate system, where the straight line perpendicular to the Z axis of the robot's key joint coordinate system intersects with the z axis of the three-dimensional vision measurement sensor coordinate system.

[0046] The position and direction of the virtual end are determined, and a new spatial motion chain model based on the virtual end is constructed.

[0047] Figure 2 is a schematic diagram of obtaining a reference coordinate system provided in an embodiment of the present invention, in which Figure 2(a) uses a circular trajectory formed by a virtual end in three-dimensional space to determine the position of a robot base coordinate system relative to a measured target coordinate system; Figure 2(b) uses a linear trajectory formed by a virtual end in three-dimensional space to determine the direction of a robot base coordinate system relative to a measured target coordinate system.

[0048] S2.1: In Figure 2(a), the robot joint J1 is the only variable of the robot and rotates to any 12 different angles. All other joint angles are kept fixed to ensure that the target is always completely within the scanning field of view of the 3D vision measurement sensor. The robot joint J1 is rotated to any 12 different angles to obtain 12 different postures.

[0049] The PnP algorithm is used to calculate the positional relationship between the measured target coordinate system and the robot base coordinate system, and the 12 trajectory points that the virtual end passes through in 12 different robot postures are recorded. These trajectory points form a circular trajectory in three-dimensional space. Through calculation and fitting, the common center of all 12 trajectory points is the center of the best fitting circle P. c (a, b, c), define γ(x i ,y i ,z i ) is the radius of the circular trajectory corresponding to the trajectory point in the robot's i-th (i=1, 2, ..., 12) posture:

[0050]

[0051] Where (x i ,y i ,z i ) is the coordinate of the i-th trajectory point. i ,y i ,z i ) and γ(x j ,y j ,z j ) is minimized and the center P of the best fitting circle is found c (a,b,c).

[0052] After finding the center of the best fitting circle Pc, set c to 0 and project it onto the xy plane of the measured target coordinate system. Therefore, (a, b, 0) is the position of the robot base coordinate system displayed in the measured target coordinate system. The z-axis of the robot base coordinate system is parallel to the z-axis of the measured target coordinate system.

[0053] S2.2: In Figure 2(b), the rotation angle of the machine joint J1 is fixed at 0°, and the basic joint J2 is rotated at any 12 different angles as the only variable of the robot. The angles of all other joints remain fixed to ensure that the target is always completely within the scanning field of view of the 3D vision measurement sensor. The robot basic joint J2 is rotated at any 12 different angles to obtain 12 different postures. The position of the virtual end in each posture is obtained by reverse calculation based on the coordinate system of the target and the image of the target captured by the 3D vision measurement sensor through the PnP algorithm.

[0054] The z component of the trajectory point coordinates obtained by the PnP algorithm is set to zero so that it is projected onto the xy plane of the target coordinate system. The best line fitting is performed on these trajectory projection points to determine the y-axis direction of the robot base coordinate system. The best fitting line is:

[0055]

[0056] Among them, (x m ,y m ) is the average center point on the best fit line containing all trajectory points, (x j ,y j ) is the coordinate of the trajectory projection point in the jth (j=1, 2, ..., 12) posture. Translate along the best fitting line to the position of the robot base coordinate system to determine the x-axis of the robot base coordinate system.

[0057] S2.3: According to the right-hand rule, the z-axis of the robot base coordinate system is obtained according to the obtained x-axis and z-axis of the robot base coordinate system. Based on the position and direction of the robot base coordinate system relative to the measured target coordinate system, the kinematic chain parameters under the spatial kinematic chain model based on the virtual end are obtained.

[0058] Figure 3 The overall flow chart provided by the embodiment of the present invention is as follows: Figure 3 As shown, the present invention provides a method for online calibration of spatial kinematic chain parameters of a robot three-dimensional vision measurement system, and the specific steps are as follows:

[0059] Complete the calibration of the measurement parameters of the 3D vision measurement sensor in the measurement system to ensure the accuracy of the subsequent measurement process. Figure 1 As shown in the figure, a new spatial motion chain model based on the virtual end is constructed and its optimization function is defined. In the third step, on this basis, the key parameters of the spatial motion chain of the measurement system are solved and optimized to realize the online calibration of the key parameters.

[0060] S3: In S2.1, the robot joint J1 rotates to any 12 different angles to obtain 12 different postures, and the 3D vision measurement sensor captures 12 different views of the target. For the robot's i-th (i=1, 2, ..., 12) posture, define is the pose transformation relationship from the 3D vision measurement sensor coordinate system to the measured target coordinate system obtained by the PnP algorithm; definition It is the position transformation relationship from the robot base coordinate system to the 3D vision measurement sensor coordinate system; define is the position transformation relationship from the robot base coordinate system to the measured target coordinate system. There is a formula:

[0061]

[0062] definition is the pose transformation relationship from the robot base coordinate system to the target coordinate system The rotational component of is the pose transformation relationship from the robot base coordinate system to the target coordinate system The translation component of R z The kinematic chain parameters from the measured target coordinate system to the robot base coordinate system The rotational component of z The kinematic chain parameters from the measured target coordinate system to the robot base coordinate system The translation component of . Using Rodrigues's rotation formula Rodrigues (R), for 12 different postures, the position error PError and orientation error OError of the spatial kinematic chain model based on the virtual end are:

[0063]

[0064] The optimization function of the model is:

[0065]

[0066] Among them, a l , d l , α l ,θ l They correspond respectively to the link length, link offset, link distortion and joint angle of the lth link in the spatial kinematic chain model based on the virtual end, where l = 1, 2, 3, 4, 5: 1-3 represent link 1, link 2, link 3 respectively; 4, 5 represent virtual arm 4', virtual arm 5' respectively.

[0067] Based on the solved optimization function in S3, the parameters of the spatial motion chain model based on the virtual end are optimized using the nonlinear iterative optimization method. The three-dimensional vision measurement sensor rescans the object to be measured to obtain the optimized data, and the optimized data is used as the initial value for re-optimization.

[0068] The maximum expected deviation threshold of the kinematic parameters of the spatial kinematic chain model of the virtual end is set to (e a , e d , e α , e θ ), and the optimization process is repeated using nonlinear iterative optimization until |a l | <e a ,|d l | <e d ,|α l | <e α ,|θ l | <e θ , the parameter optimization of the space kinematic chain model based on the virtual terminal is completed. If it cannot be optimized below the threshold, it is necessary to return to S2 to recalculate the kinematic chain parameters under the space kinematic chain model based on the virtual terminal and optimize them.

[0069] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for online calibration of spatial kinematic chain parameters of a robot three-dimensional vision measurement system, characterized in that: The steps include: Step S1: On the basis of the existing AX=XB traditional kinematics model, a new spatial kinematic chain model based on the virtual end is constructed; S2: Calculate the kinematic chain parameters of the spatial kinematic chain model based on the virtual end constructed in S1 Among them, the kinematic chain parameters refer to the rotation and translation relationship from the measured target coordinate system to the robot base coordinate system; S3: Based on the spatial kinematic chain model based on the virtual end constructed in step S1 and the kinematic chain parameters from the measured target coordinate system to the robot base coordinate system solved in step S2 As a basis, a new optimization function is constructed, and the nonlinear iterative optimization method is used to optimize the parameters of the spatial kinematic chain model based on the virtual end. The specific steps are as follows: Step S1: Construct a new spatial kinematic chain model based on the virtual end, as follows: The complete space motion chain model based on the virtual end includes the robot base, the four-axis robot link 1-link 4, the three-dimensional vision measurement sensor, the target to be measured, the virtual end and its virtual arm; the space motion chain model has the robot base coordinate system, the robot key joint coordinate system, the three-dimensional vision measurement sensor coordinate system, the target to be measured coordinate system, and the virtual end coordinate system in space; The virtual end is connected to the robot's key joints and the 3D vision measurement sensor through a virtual arm link. The position and direction are determined by the relative geometric relationship between the Z axis of the robot's key joint coordinate system and the Z axis of the 3D vision measurement sensor coordinate system, thereby completing the construction of the spatial motion chain model based on the virtual end. S2: Calculate the kinematic chain parameters of the spatial kinematic chain model based on the virtual end constructed in S1 The details are as follows: S2.1: Calculate the positional relationship from the measured target coordinate system to the robot base coordinate system The robot joint J1 is rotated as the only variable of the robot to any n different angles, n>3, and all other joint angles remain fixed to ensure that the target to be measured is always completely located in the scanning field of view of the three-dimensional vision measurement sensor; the robot joint J1 is rotated to any n different angles to obtain n different postures; based on the images of the target to be measured captured by the three-dimensional vision measurement sensor in n different postures, the positional relationship from the coordinate system of the target to be measured to the coordinate system of the robot base is calculated using the PnP algorithm; S2.2: Determine the orientation of the robot base coordinate system relative to the target coordinate system The rotation angle of robot joint J1 is fixed at 0°, and robot joint J2 is rotated to any m different angles as the only variable of the robot, m>3, and all other joint angles are kept fixed to ensure that the target to be measured is always completely located in the scanning field of view of the three-dimensional vision measurement sensor; the robot joint J2 is rotated to any m different angles to obtain m different postures; the position of the virtual end in each posture is obtained by reverse calculation based on the coordinate system of the target to be measured and the image of the target to be measured captured by the three-dimensional vision measurement sensor through the PnP algorithm; thereby determining the direction of the robot base coordinate system relative to the coordinate system of the target to be measured; S2.3: Based on the position and direction of the robot base coordinate system relative to the measured target coordinate system, the kinematic chain parameters of the spatial kinematic chain model based on the virtual end are obtained. S3: Based on the spatial kinematic chain model based on the virtual end constructed in step S1 and the kinematic chain parameters from the measured target coordinate system to the robot base coordinate system solved in step S2 As a basis, a new optimization function is constructed, and the nonlinear iterative optimization method is used to optimize the parameters of the spatial kinematic chain model based on the virtual end, as follows: In step S2.1, the robot joint J1 rotates any n different angles to obtain n different postures, and the three-dimensional visual measurement sensor captures n different views of the target to be measured; for the i-th posture of the robot, i = 1, 2, ..., n, define is the pose transformation relationship from the 3D vision measurement sensor coordinate system to the measured target coordinate system obtained by the PnP algorithm; definition It is the position transformation relationship from the robot base coordinate system to the 3D vision measurement sensor coordinate system; define is the position transformation relationship from the robot base coordinate system to the measured target coordinate system; there is a formula: definition is the pose transformation relationship from the robot base coordinate system to the target coordinate system The rotational component of is the pose transformation relationship from the robot base coordinate system to the target coordinate system The translation component of R z The kinematic chain parameters from the measured target coordinate system to the robot base coordinate system The rotational component of z The kinematic chain parameters from the measured target coordinate system to the robot base coordinate system The translation component; using Rodrigues's rotation formula Rodrigues (R), for n different postures, the position error PError and direction error OError of the spatial kinematic chain model based on the virtual end are: Then the optimization function of the spatial kinematic chain model parameters based on the virtual end is: Among them, a l , d l , α l ,θ l They correspond to the link length, link offset, link distortion and joint angle of the lth link in the space motion chain model based on the virtual end, respectively, wherein, in l=1, 2, 3, 4, 5: 1-3 represent link 1, link 2, link 3, respectively; 4, 5 represent virtual arm 4', virtual arm 5', respectively; based on the optimization function solved in step S3, after optimizing the parameters of the space motion chain model based on the virtual end using a nonlinear iterative optimization method, the three-dimensional vision measurement sensor rescans the object to be measured to obtain optimized data, and uses the optimized data as the initial value for re-optimization; The maximum expected deviation threshold of the kinematic parameters of the spatial kinematic chain model of the virtual end is set to (e a , e d , e α , e θ ), and the optimization process is repeated using nonlinear iterative optimization until |a l | <e a ,|d l | <e d ,|α l | <e α ,|θ l | <e θ , the parameter optimization of the space motion chain model based on the virtual terminal is completed; if it cannot be optimized below the threshold, it is necessary to return to step S2 to recalculate the motion chain parameters under the space motion chain model based on the virtual terminal and optimize them.

2. The method for online calibration of spatial kinematic chain parameters of a robot three-dimensional vision measurement system according to claim 1, characterized in that: In step S1, the position and direction of the virtual end are determined as follows: When the relative geometric relationship between the Z axis of the robot's key joint coordinate system and the Z axis of the 3D vision measurement sensor coordinate system is parallel, the virtual end center point position is selected at any position along the Z axis of the 3D vision measurement sensor, and the virtual end coordinate system has the same direction as the 3D vision measurement sensor coordinate system; When the relative geometric relationship between the Z axis of the robot's key joint coordinate system and the Z axis of the three-dimensional vision measurement sensor coordinate system is an intersection, the virtual end center point position is selected at the intersection of the Z axis of the three-dimensional vision measurement sensor coordinate system and the Z axis of the robot's key joint coordinate system, the Z axis of the virtual end coordinate system is along the Z axis direction of the three-dimensional vision measurement sensor coordinate system, and the X axis of the virtual end coordinate system is perpendicular to the plane formed by the two intersecting Z axes; When the relative geometric relationship between the Z axis of the robot's key joint coordinate system and the Z axis of the three-dimensional vision measurement sensor coordinate system is inclined, the virtual end center point position is selected at a point along the Z axis of the three-dimensional vision measurement sensor coordinate system, at which the straight line perpendicular to the Z axis of the robot's key joint coordinate system intersects with the z axis of the three-dimensional vision measurement sensor coordinate system.

Images