A 3D glue coating detection and analysis method based on template trajectory

Abstract

The present invention discloses a 3D glue coating detection and analysis method based on template trajectory, comprising the following steps: Step S1: Calculating the hand-eye conversion matrix, T1, T2, ..., T of each laser receiver and the end of the mobile mechanism by a line laser hand-eye calibration method; n Step S2: Pre-record template data for different glue-coated workpieces to be inspected; Step S3: In the same spatial coordinate system, adjust the poor trajectory points based on the recorded robot TCP pose and the scanned glue path point cloud; Step S4: Use the adjusted trajectory points to calculate the slice plane equations for each point; Step S5: Based on the sliced ​​data. This method reduces the amount of calculation and eliminates the need to recalculate the slice data for each run. At the same time, the data collected for each inspection is used to obtain the offset based on the positioning device and transfer the alignment to the template coordinate system. In relatively stable industrial environments, it has high reliability and more accurate segmentation.

Description

Technical Field

[0001] The present application relates to the technical field of glue coating detection, and in particular to a 3D glue coating detection and analysis method based on template trajectory. Background Art

[0002] In the current industrial field, the most common method for glue coating inspection is 2D inspection. Most 3D glue coating inspection solutions on the market are from foreign brands. These solutions usually do not calibrate the hand-eye relationship between the glue coating inspection sensor and the glue gun. Instead, they manually adjust the stitching tracks of different sensors to obtain a complete glue path contour point cloud. The glue path is analyzed by slicing the glue path, but the position and angle of the slice still rely on manually adjusting the glue path cutting center to extract the glue segment slice.

[0003] However, in the actual working process, the following problems exist when using the above solution:

[0004] 1. Without the relative relationship between the sensor and the robot, it is difficult to accurately stitch the point cloud through the robot's running trajectory to obtain accurate and precise 3D data of the glue path;

[0005] 2. The cutting of the rubber road is completely dependent on manual labor, which results in a large amount of manual intervention;

[0006] 3. The debugging process is complicated and the effect is not ideal, requiring long-term continuous adjustment. Summary of the Invention

[0007] To address the above problems, the present invention proposes a 3D glue coating detection and analysis method based on template trajectory, which includes the following steps:

[0008] Step S1: Calculate the hand-eye conversion matrix, T1, T2, ..., T of each laser receiver and the end of the mobile mechanism by using the line laser hand-eye calibration method. n ;

[0009] Step S2: pre-recording template data for different glue-coated workpieces to be inspected;

[0010] Step S3: In the same spatial coordinate system, adjust the poor trajectory points based on the recorded robot TCP pose and the scanned glue path point cloud;

[0011] Step S4: using the adjusted trajectory points, calculate the equation of the slice plane at each point;

[0012] Step S5: Analyze each segment of data based on the sliced ​​data to check the correctness of the segmentation. If it is not good, adjust the trajectory points and repeat the test until it is completely correct.

[0013] Step S6: Save the point set and its cutting plane equation as a template;

[0014] Step S7: Use the template to perform the test acquisition process, and finally realize the processing and analysis of the glue path slice data.

[0015] The further step S2 is used to record the template data, and the specific method steps are as follows:

[0016] 1) Place the normal, defect-free, glued workpiece to the inspection position;

[0017] 2) The robot calibrates the gluing TCP point pose using the 6-point method;

[0018] 3) Complete the robot gluing inspection scanning trajectory writing;

[0019] 4) Start the robot and glue detection sensor, and record the robot posture data and glue detection sensor data collected during the entire scanning process;

[0020] 5) Complete point cloud stitching based on the hand-eye relationship.

[0021] Furthermore, in step S3, the poor trajectory points are adjusted, and the specific steps are as follows:

[0022] Ⅰ) The actual running direction of the rubber road point cloud is used as the reference for the trajectory point direction, and the deviated trajectory point is considered a poor trajectory;

[0023] II) Directly adjust the coordinates of poor trajectory points to make them conform to the trend.

[0024] Furthermore, the adjustment calculation method of the trajectory point in step S3 is to adopt the following steps:

[0025] A) After collecting and recording, the original trajectory point set of the robot can be obtained, which is P ORI_1 、P ORI_2 、P ORI_3 ...P ORI_n ;

[0026] B) Observe the distribution of trajectory points on the glue coating point cloud, manually adjust the trajectory points that deviate from the glue path, and obtain the adjusted trajectory point set, which is P FIX_1 、P FIX_2 、P FIX_3 ...P FIX_n ;

[0027] C) The trajectory points are ordered points in space, according to the set chord length L chord , segment the point set and establish a set of equations for calculating piecewise quintic spline curves;

[0028] D) According to the resampling distance L resample And the obtained piecewise quintic spline curve equation group, perform point resampling, and obtain the evenly spaced slice point set P1, P2, P3...Pn .

[0029] Furthermore, the method for calculating the slice plane equations at each point in step S4 is to adopt the following steps:

[0030] Ⅰ) For the slice point set P1, P2, P3...P n For each point on the curve, take the derivative of the corresponding quintic spline curve equation to obtain the tangent normal vector N1, N2, N3...N of each point on the curve. n ;

[0031] II) According to point P n and its normal vector N n , we can get the point P n And the normal vector is N n The space plane equation A n X+B n Y+C n Z=0, A, B, C are the coefficients of the plane equation.

[0032] Furthermore, the data analysis method in step S5 adopts the following steps:

[0033] a) The overall point cloud of the rubber road at point P n At the same time, the dimensions are set as length L and height H. n 、Width W n Filter the bounding box and filter the point cloud data outside the bounding box;

[0034] b) Project the filtered data onto plane A n X+B n Y+C n At Z=0, the dimensionality of 3D data X, Y, Z is reduced to 2D data X, Y, and the processed point set is the slice data at the point.

[0035] Furthermore, the length L and height H n and width W n They are 0.5mm, 40mm and 50mm respectively.

[0036] Furthermore, the test collection process in S7 specifically includes the following steps:

[0037] 1) The workpiece to be inspected is in place, and the spatial position deviation matrix T of the workpiece relative to the template production position is calculated through four-corner positioning or 3D vision. offset ;

[0038] 2) Start the robot and glue detection sensor, and record the robot posture data and glue detection sensor data collected during the entire scanning process;

[0039] 3) According to the hand-eye relationship, complete the point cloud stitching of the glue path to be tested and obtain the point cloud PCL new ;

[0040] 4) According to the deviation matrix T offset , convert the point cloud PCL new Point P inside new_n Transformed to the corresponding point P in the template state model_n =T offset *P new_n , get the converted point cloud PCL model ;

[0041] 5) Use the established template trajectory and each point in the template trajectory and its cutting plane equation to segment the converted point cloud PCL model , get the glue path data of each frame.

[0042] Beneficial effects:

[0043] This patented solution pre-calibrates the hand-eye relationship between the robot and the glue coating detection laser sensor, and uses the hand-eye relationship to convert the stitching data, rather than manually adjusting the stitching point cloud angle corresponding to each frame after collecting the workpiece data for each project. The data stitching is more accurate, and after the calibration process is completed normally, the stitching accuracy is bound to be high.

[0044] The accuracy is not affected by the debugging personnel. Only one calibration is required to arbitrarily splice the point cloud. There is no need to adjust the splicing relationship once for each workpiece production and debugging. This solution uses preset trajectories to segment the glue path data, which can reduce the amount of calculation and does not need to recalculate the segmentation data every time it is run. At the same time, the data collected for each inspection is used to obtain the offset according to the positioning device, and the pose in the template coordinate system is transferred to align. In a relatively stable industrial environment, it has high reliability and more accurate segmentation. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 This is a template trajectory recording flow chart of a 3D glue coating detection and analysis method based on template trajectory provided by the present invention;

[0046] Figure 2 This is a flow chart of a calculation method for the entire glue path track slice in a 3D glue coating detection and analysis method based on a template track provided by the present invention;

[0047] Figure 3 This is the actual implementation of the acquisition and processing flow in the 3D glue coating detection and analysis method based on template trajectory provided by the present invention. DETAILED DESCRIPTION

[0048] In order to allow those skilled in the art to better understand the present invention, the present invention is described below in conjunction with embodiments and drawings. Figure 1-3 .

[0049] In order to achieve the content described in the present invention, the present invention designs a 3D glue coating detection and analysis method based on template trajectory, which includes the following steps:

[0050] Step S1: Calculate the hand-eye conversion matrix, T1, T2, ..., T of each laser receiver and the end of the mobile mechanism by using the line laser hand-eye calibration method. n ;

[0051] Step S2: For different glued workpieces to be inspected, template data is pre-recorded. When recording the template data, a normal, defect-free glued workpiece is first placed at the inspection position. The robot then calibrates the gluing TCP point pose using the 6-point method. The robot's gluing inspection scanning trajectory is then written. The robot and gluing inspection sensor are then started, and the robot's pose data and gluing inspection sensor data collected during the entire scanning process are recorded. Finally, point cloud stitching is completed based on the hand-eye relationship.

[0052] Step S3: In the same spatial coordinate system, according to the recorded robot TCP pose and the scanned glue path point cloud, adjust the poor trajectory points. First, use the actual glue path point cloud running direction as the trajectory point direction reference. The trajectory point with a large deviation is a poor trajectory. Then directly adjust the coordinates of the poor trajectory points to make it conform to the trend. In the same spatial coordinate system, according to the recorded robot TCP pose and the scanned glue path point cloud, fine-tune the poor trajectory points to make them approximate to the glue coating trajectory. According to the set chord length of 20mm, the point set is segmented and the calculation piecewise quintic spline curve equation group is established. According to the resampling distance of 0.1mm and the obtained piecewise quintic spline curve equation group, point resampling is performed to obtain evenly spaced slice point sets P1, P2, P3...P n ;

[0053] Step S4: Use the adjusted trajectory points to calculate the slice plane equations for each point, and calculate the slice point set P1, P2, P3...P n For each point on the curve, take the derivative of the corresponding quintic spline curve equation to obtain the tangent normal vector N1, N2, N3...N of each point on the curve. n , according to point P and its normal vector N n , we can get the point P n And the normal vector is N n The space plane equation A n X+B n Y+C n Z=0, A, B, C are the coefficients of the plane equation;

[0054] Step S5: Analyze each segment of data based on the sliced ​​data to check the correctness of the segmentation. If the segmentation is not correct, adjust the trajectory points and repeat the test until it is completely correct. nAccording to the size of the glue, a bounding box with a length of 0.5mm, a height of 40mm, and a width of 50mm is set for filtering. The point cloud data outside the bounding box is filtered and the filtered data is projected onto plane A. n X+B n Y+C n At Z=0, the dimensionality reduction from 3D data X, Y, Z to 2D data X, Y is performed, and the processed point set is the point P. n Check and analyze the template data at the slice data location and optimize and fine-tune it.

[0055] Step S6: Save the point set and its cutting plane equation as a template;

[0056] Step S7: Use the template to perform the test acquisition process, and finally realize the processing and analysis of the glue path slice data. The test acquisition process includes the following steps: the workpiece to be tested is in place, and the spatial position deviation matrix T of the workpiece with respect to the template production position is calculated by four-corner positioning or 3D vision. offset , start the robot and glue detection sensor, record the robot posture data and glue detection sensor collection data of the entire scanning process, complete the glue path point cloud splicing according to the hand-eye relationship, and obtain the point cloud PCL new , according to the deviation matrix T offset , convert the point cloud PCL new Point P inside new_n Transformed to the corresponding point P in the template state model_n =T offset *P new_n , get the converted point cloud PCL model , using the established template trajectory, using each point in the template trajectory and its cutting plane equation, segment the converted point cloud PCL model , get the glue path data of each frame.

[0057] In summary, this solution uses preset trajectories to segment the glue road data, which can reduce the amount of calculation and eliminate the need to recalculate the segmentation data every time it is run. At the same time, the data collected for each detection is transferred to the alignment template coordinate system based on the offset obtained by the positioning device. This has high reliability and more accurate segmentation in a relatively stable industrial environment.

Claims

1. A 3D glue detection and analysis method based on template trajectory, characterized in that: The following steps are included: Step S1: Calculate the hand-eye conversion matrix T1, T2, ..., T between each laser receiver and the end of the mobile mechanism by using the line laser hand-eye calibration method. n ; Step S2: pre-recording template data for different glue-coated workpieces to be inspected; Step S3: In the same spatial coordinate system, adjust the poor trajectory points based on the recorded robot TCP pose and the scanned glue path point cloud; Step S4: using the adjusted trajectory points, calculate the equation of the slice plane at each point; Step S5: Analyze each segment of data based on the sliced ​​data to check the correctness of the segmentation. If it is not good, adjust the trajectory points and repeat the test until it is completely correct. Step S6: Save the point set and its cutting plane equation as a template; Step S7: Use the template to perform the test acquisition process, and finally realize the processing and analysis of the glue path slice data.

2. The 3D glue coating detection and analysis method based on template trajectory according to claim 1 is characterized in that: The step S2 is used to record the template data, and the specific method steps are as follows: 1) Place the normal, defect-free, glued workpiece to the inspection position; 2) The robot calibrates the gluing TCP point pose using the 6-point method; 3) Complete the robot gluing inspection scanning trajectory writing; 4) Start the robot and glue detection sensor, and record the robot posture data and glue detection sensor data collected during the entire scanning process; 5) Complete point cloud stitching based on the hand-eye relationship.

3. The 3D glue coating detection and analysis method based on template trajectory according to claim 2 is characterized in that: The specific method and steps for adjusting the poor trajectory points in step S3 are as follows: Ⅰ) The actual running direction of the rubber road point cloud is used as the reference for the trajectory point direction, and the deviated trajectory point is considered a poor trajectory; II) Directly adjust the coordinates of the poor trajectory points to make them conform to the trend.

4. The 3D glue coating detection and analysis method based on template trajectory according to claim 3 is characterized in that: The method for calculating and adjusting the trajectory points in step S3 is to use the following steps: A) After collecting and recording, the original trajectory point set of the robot can be obtained, which is P ORI_1 、P ORI_2 、P ORI_3 ...P ORI_n ; B) Observe the distribution of trajectory points on the glue coating point cloud, manually adjust the trajectory points that deviate from the glue path, and obtain the adjusted trajectory point set, which is P FIX_1 、P FIX_2 、P FIX_3 ...P FIX_n ; C) The trajectory points are ordered points in space, according to the set chord length L chord , segment the point set and establish a set of equations for calculating piecewise quintic spline curves; D) According to the resampling distance L resample And the obtained piecewise quintic spline curve equation group, perform point resampling, and obtain the evenly spaced slice point set P1, P2, P3...P n .

5. The 3D glue coating detection and analysis method based on template trajectory according to claim 4 is characterized in that: The method for calculating the plane equation of each cutting point in step S4 is to use the following steps: I) For the slice point set P1, P2, P3...P n For each point on the curve, take the derivative of the corresponding quintic spline curve equation to obtain the tangent normal vector N1, N2, N3...N of each point on the curve. n ; II) According to point P n and its normal vector N n , we can get the point P n And the normal vector is N n The space plane equation A n X+B n Y+C n Z=0,A n 、B n 、C n are the coefficients of the plane equation.

6. The 3D glue coating detection and analysis method based on template trajectory according to claim 5 is characterized in that: The data analysis method in step S5 adopts the following steps: a) The overall point cloud of the rubber road at point P n At the same time, the dimensions are set as length L and height H. n 、Width W n Filter the bounding box and filter the point cloud data outside the bounding box; b) Project the filtered data onto plane A n X+B n Y+C n At Z=0, the dimensionality of 3D data X, Y, Z is reduced to 2D data X, Y, and the processed point set is the slice data at the point.

7. The 3D glue coating detection and analysis method based on template trajectory according to claim 6 is characterized in that: The length L and height H n and width W n They are 0.5mm, 40mm and 50mm respectively.

8. The 3D glue coating detection and analysis method based on template trajectory according to claim 7 is characterized in that: The test collection process in S7 is specifically as follows: 1) The workpiece to be inspected is in place, and the spatial position deviation matrix T of the workpiece relative to the template production position is calculated through four-corner positioning or 3D vision. offset ; 2) Start the robot and glue detection sensor, and record the robot posture data and glue detection sensor data collected during the entire scanning process; 3) According to the hand-eye relationship, complete the point cloud stitching of the glue path to be tested and obtain the point cloud PCL new ; 4) According to the deviation matrix T offset , convert the point cloud PCL new Point P inside new_n Transformed to the corresponding point P in the template state model_n =T offset *P new_n , get the converted point cloud PCL model ; 5) Use the established template trajectory and each point in the template trajectory and its cutting plane equation to segment the converted point cloud PCL model , get the glue path data of each frame.

Images