A multi-point laser positioning method for copper plate stack of cathode copper packaging production line

Abstract

This invention discloses a multi-point laser positioning method for copper plate stacks in a cathode copper packaging production line. An industrial robot is positioned at the labeling station of the copper plate finished product packaging production line, and a laser displacement sensor is installed on the side of the robot's end effector. When the copper plate stack enters the labeling station along with the finished copper plate packaging production line, the robot, carrying the actuator, moves above the copper plate stack. The actuator collects data points along an elliptical trajectory in the horizontal direction. After data collection, the obtained measurement data is first transformed to obtain the copper plate surface coordinate data in the tool coordinate system. Then, multiple sets of normal vectors are obtained using the coordinate data combined with the combination principle and cross product method. Finally, the optimal vector for the plane is obtained using cluster analysis. This method overcomes the shortcomings of traditional single-plane fitting methods, which are susceptible to local defects and measurement noise interference. It significantly improves the accuracy and robustness of the copper plate stack's attitude positioning, providing a reliable basis for subsequent precise labeling by the robot.

Description

Technical Field

[0001] This invention belongs to the field of multi-point laser positioning methods for copper plate stacks, and particularly relates to a multi-point laser positioning method for copper plate stacks used in cathode copper packaging production lines. Background Technology

[0002] Copper is widely used in people's production and daily life and is a very important heavy metal material. It is generally processed into copper plates through production lines and then stacked into copper plate stacks according to demand.

[0003] In the packaging production line of finished cathode copper, labels need to be affixed to the surface of copper plate stacks. Existing technology typically uses laser displacement sensors mounted on the end effector of industrial robots to acquire surface data of the copper plates by scanning along a specific trajectory, and then estimates the surface normal vector to adjust the robot's posture. Elliptical trajectory sampling is widely used due to its ability to cover a large area and its smooth motion. However, existing methods often use least squares or principal component analysis to directly fit a plane after acquiring the data. These global fitting methods are extremely sensitive to measurement noise and local defects on the copper plate surface (such as bumps, pits, oxide scale, etc.). Once the scanning path passes through a defect area, the fitted normal vector will have a significant deviation, leading to skewed labeling angles, label detachment, or even robot collision risks. Therefore, how to effectively suppress abnormal data interference while retaining the advantages of elliptical trajectory sampling and improving the accuracy and robustness of normal vector estimation has become a pressing technical problem to be solved in this field. Summary of the Invention

[0004] (1) Technical problem to be solved: Traditional fitting methods are extremely sensitive to measurement noise and local defects on the copper plate surface. Once the scanning path passes through the defect area, the fitted normal vector will have a significant deviation, resulting in skewed labeling angle, label falling off, or even robot collision risk. This invention provides a multi-point laser positioning method for copper plate stacks in cathode copper packaging production line. It does not require changing the original equipment scheme. By introducing a clustering solution strategy of multiple sets of normal vectors, consensus direction is extracted from massive point pairs, thereby obtaining a stable and reliable optimal normal vector. The attitude positioning of the copper plate stack surface on the production line can be completed conveniently and quickly.

[0005] (2) The technical solution adopted in this invention is as follows: A multi-point laser positioning method for copper plate stacks in a cathode copper packaging production line, comprising: 1. Elliptical trajectory point acquisition: The robot, carrying an actuator equipped with a laser displacement sensor, moves to the top of the copper plate stack. The actuator collects data points along an elliptical trajectory in the horizontal direction, collecting distance data from multiple laser emission windows to the copper plate surface. This data is then combined with the robot's pose data and transformed to obtain a set of three-dimensional coordinate points on the copper plate surface in the tool coordinate system. By fusing the robot's pose data, the distance values ​​are converted into a set of three-dimensional coordinate points P={p1,p2,…,pm} in the tool coordinate system.

[0006] 2. Based on the principle of combination, select all non-collinear three-point combinations from the set of three-dimensional coordinate points, and calculate the normal vector corresponding to each combination by cross product method to obtain the initial set of normal vectors; 3. The principal component analysis method is used to calculate the principal direction of the three-dimensional coordinate point set, obtain the reference normal vector, and calculate the angle between each vector in the initial normal vector set and the reference normal vector, and remove gross error vectors whose angle exceeds three times the standard deviation. Fourth, the minimum-maximum method is used to select the initial cluster centers for the remaining set of normal vectors. K-Means clustering analysis is then performed, and the class with the largest number of samples is selected. The mean of the normal vectors of this class is then calculated as the optimal normal vector for the copper plate surface. 5. Input the optimal normal vector into the production line control system. The industrial robot adjusts the actuator posture according to the vector to complete the labeling process.

[0007] A further technical solution is as follows: In step two, the non-collinear three-point combinations are filtered by excluding continuous or approximately collinear points. Using the principle of combination, all non-collinear three-point combinations (excluding continuous or approximately collinear points) are selected from the point set P. For each valid combination (pi, pj, pk), the plane normal vector n determined by the combination is calculated by the vector cross product. ijk =(pj−pi)×(pk−pi), and normalize to obtain the initial set of normal vectors N={n1,n2,…,nK}.

[0008] A further technical solution is as follows: In step three, principal component analysis is used to perform eigenvalue decomposition of the covariance matrix of the original point set P, and the eigenvector corresponding to the smallest eigenvalue is taken as the reference normal vector n. pca Calculate each n l ∈N and n pca The included angle θ l =arccos(∣n l ·n pca ∣), calculate the standard deviation σ of the set of included angles, and discard those that satisfy ∣θ l The vectors with -θˉ∣>3σ are used to obtain the set of effective normal vectors N′.

[0009] A further technical solution involves: in step four, selecting two initial cluster centers for N′ using the minimax method, and then performing K-Means clustering. After clustering, the class with the largest number of samples is selected, and the arithmetic mean of all vectors in that class is calculated as the optimal normal vector n for the copper plate surface. opt .

[0010] (3) Due to the adoption of the above technical solution, the beneficial effects of the present invention are: 1. An industrial robot is placed at the labeling station of the copper plate finished product packaging production line, and a laser displacement sensor is installed on the side of the robot's end effector. When the copper plate stack enters the labeling station along the copper plate finished product packaging production line, the robot carries the actuator to move above the copper plate stack. The actuator collects data points along an elliptical trajectory in the horizontal direction. After the data collection is completed, the obtained measurement data is first transformed to obtain the copper plate surface coordinate data in the tool coordinate system. Then, the coordinate data is combined with the combination principle and cross product method to obtain multiple sets of normal vectors. Finally, the optimal vector in the plane is obtained by cluster analysis. This method overcomes the shortcomings of the traditional single plane fitting method, which is susceptible to local defects and measurement noise interference. It significantly improves the accuracy and robustness of the copper plate stack attitude positioning, and provides a reliable basis for the robot to accurately label the copper plates in the future.

[0011] 2. Control the robotic arm carrying a laser displacement sensor to collect data on the copper plate surface along an elliptical trajectory in the horizontal direction. This avoids measurement errors caused by surface defects in the copper plate, greatly improving the accuracy and continuity of the measurement data.

[0012] 3. By utilizing the combinatorial principle to select different combinations of points on the plane, and after eliminating unreasonable combinations, the vector sets of data in different combinations are calculated. Then, the minimum-maximum method is used to select the initial cluster centers for K-Means clustering to obtain the optimal vector. This algorithm can effectively avoid problems such as the failure of working surface normal vector estimation due to measurement errors. Attached Figure Description

[0013] Figure 1 A schematic diagram showing the positional relationship between the robot and the copper plate stack in this invention is shown; Figure 2 It shows Figure 1 A schematic diagram of the multi-point positioning method for copper plates, showing the points taken along an elliptical trajectory on the horizontal plane. Figure 3 The flowchart of the optimal vector algorithm for multi-point positioning of copper plates by combining multiple normal vectors and clustering is shown. Figure 4 This demonstrates the conversion from the laser measurement coordinate system to the robot tool coordinate system; Figure 5 The multi-point normal vector estimation method is shown in the simulated plane data diagram; Figure 6 The graph shows the measurement data with added errors. Detailed Implementation

[0014] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0015] like Figures 1-6 As shown.

[0016] Example 1: A multi-point laser positioning method for copper plate stacks in a cathode copper packaging production line, including... 1. Elliptical trajectory point acquisition: The robot carries the actuator equipped with a laser displacement sensor and moves to the top of the copper plate stack. The actuator collects data points along the elliptical trajectory in the horizontal direction, collecting distance data from multiple laser emission windows to the copper plate surface. Combined with the robot pose data, the coordinate transformation is performed to obtain the three-dimensional coordinate point set of the copper plate surface in the tool coordinate system. 2. Based on the principle of combination, select all non-collinear three-point combinations from the set of three-dimensional coordinate points, and calculate the normal vector corresponding to each combination by cross product method to obtain the initial set of normal vectors; 3. The principal component analysis method is used to calculate the principal direction of the three-dimensional coordinate point set, obtain the reference normal vector, and calculate the angle between each vector in the initial normal vector set and the reference normal vector, and remove gross error vectors whose angle exceeds three times the standard deviation. Fourth, the minimum-maximum method is used to select the initial cluster centers for the remaining set of normal vectors. K-Means clustering analysis is then performed, and the class with the largest number of samples is selected. The mean of the normal vectors of this class is then calculated as the optimal normal vector for the copper plate surface. 5. Input the optimal normal vector into the production line control system. The industrial robot adjusts the actuator posture according to the vector to complete the labeling process.

[0017] A further technical solution is as follows: In step two, the non-collinear three-point combinations are filtered by excluding continuous or approximately collinear points. Using the principle of combination, all non-collinear three-point combinations (excluding continuous or approximately collinear points) are selected from the point set P. For each valid combination (pi, pj, pk), the plane normal vector n determined by the combination is calculated by the vector cross product. ijk =(pj−pi)×(pk−pi), and normalize to obtain the initial set of normal vectors N={n1,n2,…,nK}.

[0018] In step three, principal component analysis is used to perform eigenvalue decomposition of the covariance matrix of the original point set P, and the eigenvector corresponding to the smallest eigenvalue is taken as the reference normal vector n. pca Calculate each n l ∈N and n pcaThe included angle θ l =arccos(∣n l ·n pca ∣), calculate the standard deviation σ of the set of included angles, and discard those that satisfy ∣θ l The vectors with -θˉ∣>3σ are used to obtain the set of effective normal vectors N′.

[0019] In step four, the minimax method is used to select two initial cluster centers for N′, and K-Means clustering is performed. After clustering, the class with the largest number of samples is selected, and the arithmetic mean of all vectors in that class is calculated as the optimal normal vector n of the copper plate surface. opt .

[0020] An industrial robot is positioned at the labeling station of copper plate packaging production line 1. A laser displacement sensor is installed on the side of the actuator, with the laser emission window parallel to the actuator's working surface. When the copper plate stack 2 enters the labeling station along the finished copper plate packaging production line, the industrial robot, carrying the actuator, moves above the surface of the copper plate stack. The laser displacement sensor then activates, and the actuator moves horizontally along an elliptical trajectory. At this point, the laser displacement sensor measures only the distance from the laser emission window to the copper plate surface. By incorporating the industrial robot's actuator pose data and performing coordinate transformation, the copper plate surface data in the tool coordinate system is obtained. After measurement, multiple sets of normal vectors are calculated based on the combination principle, and then a clustering algorithm is used to obtain the optimal vector for the copper plate surface. The vector information is then transmitted to the system, and the industrial robot adjusts its pose to proceed to the next process.

[0021] This invention utilizes the principle of combination to select planar data, obtaining combinations of different points. After eliminating unreasonable combinations, it calculates the vector set of data in each combination. The unreasonable combinations include combinations of three points that are approximately collinear. The criterion for determining whether three points are approximately collinear is: calculate the area S of the triangle formed by the three points A, B, and C. If S is less than a preset threshold T (e.g., T = 0.01 mm)... 2 (), or the included angle between any two sides of the three points is less than 5 ohms. ∘ or greater than 175 ∘ If the three points are approximately collinear, then they are considered to be approximately collinear. Since three approximately collinear points cannot stably define a plane (a small measurement error can cause a drastic change in the direction of the normal vector), such combinations will be discarded and will not participate in the normal vector calculation. Then, the minimax method is used to select initial cluster centers for K-Means clustering to obtain the optimal vector. This algorithm effectively avoids problems such as the failure of working surface normal vector estimation due to measurement errors.

[0022] For example: Figure 1As shown, industrial robots 3 and 4 are installed at the labeling station of the cathode copper packaging production line 1. A laser displacement sensor 6 is fixed to the side of their end effector 5, with the laser emission direction parallel to the working surface of the actuator. When the copper plate stack 2 arrives at the labeling station along the production line, the robot moves the actuator to a position approximately 200mm above the surface of the copper plate stack. The actuator moves along the horizontal plane 7... Figure 2 The elliptical trajectory shown has a major-to-minor axis ratio of approximately 2:1, ensuring coverage of the main area of ​​the copper plate surface. During the motion, the laser displacement sensor continuously acquires distance data d at a sampling frequency of 1kHz. i Simultaneously, the angles of each joint of the robot are recorded, and the actuator pose is calculated through forward kinematics. Let the transformation matrix of the laser sensor coordinate system relative to the tool coordinate system be... Then the coordinates of the measurement point in the tool coordinate system are: The coordinates are then further transformed to a base coordinate system to obtain a three-dimensional point set P. In this embodiment, a total of 9 trajectory points were collected.

[0023] Example 2: Solving clustering problems using multiple normal vector combinations Taking the nine collected points as an example, such as Figure 3 As shown, the algorithm flow is as follows: (1) Generate normal vectors by combining three points: Select all non-collinear combinations from the 9 points. There are a total of C3 + 9 = 84 combinations from the 9 points, but the cases where the three points are continuous or approximately collinear (e.g., combinations formed by adjacent trajectory points) need to be excluded. Calculate the area of ​​the triangle formed by the three points. If the area is less than the threshold, it is considered collinear and discarded. The remaining valid combinations are assumed to be 60 groups. For each group of three points A, B, C, calculate the normal vector: The initial set of normal vectors is obtained as N = {n1, ..., n}. 60}

[0024] (2) Removal of gross errors: Principal component analysis was performed on the original 9 points to construct the covariance matrix. ,in The center of gravity. To Perform eigenvalue decomposition; the eigenvector corresponding to the smallest eigenvalue is the reference normal vector n. pca Calculate the relationship between each vector in N and n. pca The included angle θ l The mean and standard deviation of the included angle σ are obtained. |θ is then discarded. l For vectors with a value of −∣>3σ, the remaining valid vector set is N′ (e.g., 52 groups).

[0025] (3) K-Means clustering optimization: Apply the minimax method to N′ to select two initial cluster centers. First, calculate the sum of the Euclidean distances of each vector to other vectors, and take the largest sum as the first center C1; then calculate the distances of the remaining vectors to C1, and take the largest distance as the second center C2. Perform K-Means iteration with C1 and C2 as the initial centers until the centers converge. Count the number of samples in the two clusters, take the cluster with more samples (e.g., 40 groups), calculate the mean of all vectors in the cluster, and obtain the optimal normal vector n. opt .

[0026] Example 3: Effect Verification To verify the effectiveness of this method, an ideal plane x+y+z=1 was generated in a simulation environment, with the true value of the normal vector being (0.5773, 0.5773, 0.5773). Nine points were randomly selected from the plane, and Z-axis noise with a mean of 0 and a standard deviation of 0.05 was added to simulate measurement error. The normal vector was estimated using the least squares method, principal component analysis, and the proposed method, and the results are compared below: method Estimate the normal vector (a, b, c) Angle error between the true value and the actual value (°) Least squares method (0.5841, 0.5871, 0.5706) 1.52 Principal component analysis (0.5821, 0.5835, 0.5882) 1.48 This method (0.5741, 0.5726, 0.5782) 0.63 Table 1 It is evident that this method exhibits the smallest error and the strongest noise resistance.

[0027] In actual copper plate surface testing, data was collected using an OPTEX CD33-250N laser displacement sensor, and the stability of the normal vector before and after correction was compared. Table 2 shows the results of direct fitting of the original data and the results after correction using this method in two measurements: Serial Number Handling method Measured value variance α / degree β / degree γ / degree 1 Before revision 735.00 91.4026 1.4026 107.1567 Revised 28.30 91.8749 1.9806 108.0204 2 Before revision 671.21 90.8305 36.2542 110.6627 Revised 45.10 91.4501 37.1151 109.4685 Table 2 The correction significantly reduced the fluctuation of the normal vector, indicating that the method effectively suppressed the interference of measurement noise and surface defects.

[0028] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Any simple modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A multi-point laser positioning method for copper plate stacks in a cathode copper packaging production line, characterized in that: include 1. Elliptical trajectory point acquisition: The robot carries the actuator equipped with a laser displacement sensor and moves to the top of the copper plate stack. The actuator collects data points along the elliptical trajectory in the horizontal direction, collecting distance data from multiple laser emission windows to the copper plate surface. Combined with the robot pose data, the coordinate transformation is performed to obtain the three-dimensional coordinate point set of the copper plate surface in the tool coordinate system.

2. Based on the principle of combination, select all non-collinear three-point combinations from the set of three-dimensional coordinate points, calculate the normal vector corresponding to each combination, and obtain the initial set of normal vectors.

3. The principal component analysis method is used to calculate the principal direction of the three-dimensional coordinate point set, obtain the reference normal vector, and calculate the angle between each vector in the initial normal vector set and the reference normal vector, and remove gross error vectors whose angle exceeds three times the standard deviation. Fourth, the minimum-maximum method is used to select the initial cluster centers for the remaining set of normal vectors. K-Means clustering analysis is then performed, and the class with the largest number of samples is selected. The mean of the normal vectors of this class is then calculated as the optimal normal vector for the copper plate surface.

5. Input the optimal normal vector into the production line control system. The industrial robot adjusts the actuator posture according to the vector to complete the labeling process.

2. The multi-point laser positioning method for copper plate stacks in a cathode copper packaging production line according to claim 1, characterized in that, In step two, the non-collinear three-point combinations are filtered by excluding continuous or approximately collinear points. Using the principle of combination, all non-collinear three-point combinations are selected from the point set P (excluding continuous or approximately collinear points). For each valid combination (pi, pj, pk), the plane normal vector n determined by the combination is calculated using the vector cross product. ijk =(pj−pi)×(pk−pi), and normalize to obtain the initial set of normal vectors N={n1,n2,…,nK}.

3. The multi-point laser positioning method for copper plate stacks in a cathode copper packaging production line according to claim 2, characterized in that: In step three, principal component analysis is used to perform eigenvalue decomposition of the covariance matrix of the original point set P, and the eigenvector corresponding to the smallest eigenvalue is taken as the reference normal vector n. pca Calculate each n l ∈N and n pca The included angle θ l =arccos(∣n l ·n pca ∣), calculate the standard deviation σ of the set of included angles, and discard those that satisfy ∣θ l The vectors with -θˉ∣>3σ are used to obtain the set of effective normal vectors N′.

4. The multi-point laser positioning method for copper plate stacks in a cathode copper packaging production line according to claim 3, characterized in that: In step four, the minimax method is used to select two initial cluster centers for N′, and K-Means clustering is performed. After clustering, the class with the largest number of samples is selected, and the arithmetic mean of all vectors in that class is calculated as the optimal normal vector n of the copper plate surface. opt .

Images