Aluminum ingot labeling position identification method based on 3D vision
Through the aluminum ingot labeling position recognition method based on 3D vision, the shortcomings of the existing system in accurately identifying and positioning the labeling position are solved, efficient and accurate labeling operations are achieved, safety and adaptability are enhanced, and production costs are reduced.
Patent Information
- Application Number
- CN202510112374.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-23
- Publication Date
- 2025-06-06
AI Technical Summary
The existing aluminum ingot labeling system has shortcomings in accurately identifying and positioning the labeling position, resulting in low labeling efficiency, low accuracy and safety hazards.
The aluminum ingot labeling position recognition method based on 3D vision is used to obtain the point cloud data on the surface of the aluminum ingot, perform point cloud preprocessing and rectangle fitting, and the accurate labeling position is calculated and sent to the robot for labeling.
It improves the efficiency and accuracy of the aluminum ingot labeling process, reduces manual intervention, enhances the safety of operation, strong adaptability, reduces production costs, and improves intelligence.
Smart Images
Figure CN120107351A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of industrial automation technology, and in particular to a method for identifying the labeling position of an aluminum ingot based on 3D vision. Background Art
[0002] In modern manufacturing, aluminum ingots usually need to be labeled to identify product information, such as specifications, batch numbers, production dates, etc. However, the traditional labeling process usually relies on manual operation, which is not only inefficient but also prone to human errors. In addition, manual labeling poses a great safety hazard, especially when the operating environment is complex or large quantities of aluminum ingots need to be processed. With the development of industrial automation technology, the use of 3D vision systems combined with automated equipment for labeling has become a feasible solution, but the existing system still has shortcomings in accurately identifying and locating the labeling position. Summary of the invention
[0003] The purpose of the present invention is to overcome the shortcomings of the prior art, avoid the low efficiency and inaccurate labeling position of aluminum ingots, and provide a method for identifying the labeling position of aluminum ingots based on 3D vision. This method can improve the efficiency of the aluminum ingot labeling process and reduce manual intervention; improve labeling accuracy and ensure the accuracy of label position; enhance operational safety and reduce safety hazards caused by human factors.
[0004] The object of the present invention is achieved by the following measures: A method for identifying the position of aluminum ingot labeling based on 3D vision, comprising the following steps:
[0005] S1: Obtain point cloud data. The camera installed on the robot arm is aimed at the aluminum ingot to be labeled to take pictures and scan, and obtain the point cloud data of the aluminum ingot surface;
[0006] S2: performing point cloud preprocessing on the acquired aluminum ingot surface point cloud data to filter out point cloud noise;
[0007] S3: After the noise is filtered out, the point cloud is fitted into the shape of the aluminum ingot according to a rectangle, and the labeling position is calculated according to the corner points of the fixed position of the fitted rectangle;
[0008] S4: The labeling position calculated based on 3D vision is sent to the robot, and the robot labels the target aluminum ingot.
[0009] Preferably, in the step S2, the point cloud preprocessing uses a filtering algorithm to perform noise reduction processing, and the filtering algorithm includes voxel grid sampling, side wall filtering, statistical denoising, and region segmentation.
[0010] Preferably, the voxel grid sampling method is as follows:
[0011] The point cloud is divided into three-dimensional grids, each grid is represented by a representative point, which is used to reduce the density of point cloud data and simplify the data. The formula is:
[0012]
[0013] Among them, p i is the point in the grid and N is the number of points in the grid.
[0014] Preferably, the sidewall filtering method is as follows:
[0015] Side wall filtering mainly uses normal vector analysis to calculate the normal vector of each point and determine whether it belongs to the side wall based on the direction of the normal vector. If the angle between the normal vector and the horizontal plane exceeds a certain threshold, it is considered to be a side wall point and filtered. The formula is as follows:
[0016]
[0017] Where n is the normal vector of the point and v is the reference direction vector.
[0018] Preferably, the statistical denoising method is as follows:
[0019] The average distance between each point and its neighboring points is calculated. If the distance exceeds a certain statistical threshold, the point is considered an outlier. The formula is:
[0020]
[0021] threshold=μ + ασ
[0022] Among them, p is the point before w, p i are k neighboring points, μ is the average distance of all points, σ is the standard deviation, and α is the adjustment parameter.
[0023] Preferably, the region segmentation method is as follows:
[0024] Use Euclidean distance to measure the distance between two points in space. 1 ,y 1 ,z 1 ) and q=(x 2 ,y 2 ,z 2 ), the Euclidean distance is defined as:
[0025]
[0026] This formula is used to calculate the straight-line distance between points to determine their proximity in space. In the Euclidean distance segmentation method, the Euclidean clustering extraction algorithm is used.
[0027] Preferably, the rectangle fitting process of step S3 includes performing principal component analysis on the aluminum ingot point cloud, determining the main direction of the point cloud by calculating the covariance matrix of the point cloud data, and realizing the fitting of the minimum circumscribed rectangle, the steps are:
[0028] S3a: Calculate the point cloud centroid
[0029]
[0030] Among them, p i is the i-th point in the point cloud, N is the total number of points, and c is the centroid;
[0031] S3b: Calculate the covariance matrix
[0032]
[0033] S3c: Calculate the eigenvalues and eigenvectors of the covariance matrix. The eigenvectors correspond to the main directions of the point cloud. Arrange the eigenvalues and select the eigenvectors corresponding to the two largest eigenvalues as the main directions.
[0034] S3d: Rectangle fitting, projects the point cloud in two main directions and calculates the minimum bounding box of the projection points to obtain the four vertices of the rectangle. By adding a tool offset to one of the corner points of the rectangle, the labeling position of the aluminum ingot can be obtained by moving along the tool.
[0035] Compared with the prior art, the present invention has the following beneficial effects:
[0036] 1. Improve labeling efficiency:
[0037] The 3D vision system can automatically identify and locate the labeling position of aluminum ingots, greatly reducing the time of manual participation and achieving efficient labeling. Compared with traditional manual operation, this method can greatly increase the labeling speed and meet the needs of mass production.
[0038] 2. Enhance labeling accuracy:
[0039] Using precise 3D point cloud data analysis and fitting technology, the geometric shape and labeling position of the aluminum ingot can be accurately calculated to ensure that the label is affixed to the specified position and reduce poor labeling caused by errors.
[0040] 3. Improved security:
[0041] By replacing manual labeling operations with automated equipment, the operating risks of workers in complex environments are reduced, the overall safety of the production line is improved, and labeling errors caused by human error are reduced.
[0042] 4. Strong adaptability:
[0043] The 3D vision system of the present invention can adapt to aluminum ingots of different shapes and specifications, and can achieve efficient processing of diversified products by flexibly adjusting the recognition algorithm and labeling position, and has broad application prospects.
[0044] 5. Reduce production costs:
[0045] This method can reduce the company's production costs by reducing reliance on manual labor and improving operational efficiency. In addition, it can reduce rework and quality losses caused by labeling errors and further optimize the production process.
[0046] 6. High degree of intelligence:
[0047] This method makes full use of 3D vision technology and automated mechanical equipment, embodies a high level of intelligence and automation capabilities, conforms to the development direction of Industry 4.0, and promotes the transformation of traditional industries to intelligent manufacturing. BRIEF DESCRIPTION OF THE DRAWINGS
[0048] Figure 1 This is a workflow diagram for aluminum ingot labeling based on 3D vision. DETAILED DESCRIPTION
[0049] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only 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.
[0050] Example 1: Figure 1 As shown, a method for identifying the position of aluminum ingot labeling based on 3D vision includes the following steps:
[0051] S1: Obtain point cloud data. The camera installed on the robot arm is aimed at the aluminum ingot that needs to be labeled to take pictures and scan, and obtain the surface point cloud data of the aluminum ingot. Specifically, 1. Calibrate the camera and sensor. Use the calibration plate to calibrate the camera to obtain the internal and external parameters of the camera to ensure the consistency of the image with the real space. Align the camera coordinate system with the robot coordinate system through hand-eye calibration (Hand-EyeCalibration). 2. Robot positioning and shooting. Let the robot arm move the camera to the preset fixed shooting position to ensure that all aluminum ingots that need to be labeled are included in the field of view. Use a 3D vision sensor (such as LiDAR or stereo camera) to scan the aluminum ingot to obtain surface point cloud data.
[0052] S2: Perform point cloud preprocessing on the acquired aluminum ingot surface point cloud data to filter out point cloud noise; that is, point cloud noise that interferes with fitting positioning, such as the side wall of the aluminum ingot, the aluminum ingot block that does not need to be labeled, and the binding rope, needs to be filtered out.
[0053] S3: After the noise is filtered out, the point cloud is fitted into the shape of the aluminum ingot according to a rectangle, and the labeling position is calculated according to the corner points of the fixed position of the fitted rectangle;
[0054] S4: The labeling position calculated based on 3D vision is sent to the robot, and the robot labels the target aluminum ingot.
[0055] In step S2, point cloud preprocessing uses a filtering algorithm to perform noise reduction to remove unnecessary background noise and isolated points, thereby improving the quality and accuracy of point cloud data; the filtering algorithm includes voxel grid sampling, side wall filtering, statistical denoising, and region segmentation.
[0056] Voxel Grid Sampling simplifies the data by reducing the number of points while maintaining the overall geometric features, which helps speed up subsequent processing steps and reduce computational complexity. The method is as follows:
[0057] The point cloud is divided into three-dimensional grids, and each grid is represented by a representative point (usually the centroid of the grid point or a randomly selected point) to reduce the density of the point cloud data and simplify the data. The formula is:
[0058]
[0059] Among them, p i is the point in the grid and N is the number of points in the grid.
[0060] Side wall filtering is used to remove points related to the side wall of the aluminum ingot, which are usually not required to participate in the calculation of the labeling position. The method is as follows:
[0061] Side wall filtering mainly uses normal vector analysis to calculate the normal vector of each point and determine whether it belongs to the side wall based on the direction of the normal vector. If the angle between the normal vector and the horizontal plane (or the expected aluminum ingot surface) exceeds a certain threshold, it is considered to be a side wall point and filtered. The formula is as follows:
[0062]
[0063] Where n is the normal vector of the point and v is the reference direction vector (eg vertical direction).
[0064] Statistical denoising is a method to remove isolated points and outliers. It relies on statistical characteristics to determine the abnormality of points. The method is as follows:
[0065] For each point, the average distance between it and the adjacent points is calculated. If the distance exceeds a certain statistical threshold (such as the mean plus or minus the standard deviation), the point is considered an outlier. The formula is:
[0066]
[0067] threshold=μ+ασ
[0068] Among them, p is the point before w, p i are k neighboring points, μ is the average distance of all points, σ is the standard deviation, and α is the adjustment parameter.
[0069] Region segmentation is used to segment the point cloud into multiple regions with independent features for further processing. The method is as follows:
[0070] Use Euclidean distance to measure the distance between two points in space. 1 ,y 1 ,z 1 ) and q=(x 2 ,y 2 ,z 2 ), the Euclidean distance is defined as:
[0071]
[0072] This formula is used to calculate the straight-line distance between points to determine their proximity in space. In the Euclidean distance segmentation method, the Euclidean clustering extraction algorithm is used.
[0073] The rectangle fitting process of step S3 includes performing principal component analysis (PCA) on the aluminum ingot point cloud, determining the main direction of the point cloud by calculating the covariance matrix of the point cloud data, and realizing the fitting of the minimum circumscribed rectangle. The steps are as follows:
[0074] S3a: Calculate the point cloud centroid
[0075]
[0076] Among them, p i is the i-th point in the point cloud, N is the total number of points, and c is the centroid;
[0077] S3b: Calculate the covariance matrix
[0078]
[0079] S3c: Calculate the eigenvalues and eigenvectors of the covariance matrix. The eigenvectors correspond to the main directions of the point cloud. Arrange the eigenvalues and select the eigenvectors corresponding to the two largest eigenvalues as the main directions.
[0080] S3d: Rectangle fitting, projects the point cloud in two main directions and calculates the minimum bounding box of the projection points to obtain the four vertices of the rectangle. By adding a tool offset to one of the corner points of the rectangle, the labeling position of the aluminum ingot can be obtained by moving along the tool.
[0081] Guide the robot to label, control the robot to move to the labeling position, and ensure that the labeling device is perpendicular to the surface of the aluminum ingot. Perform the labeling operation to ensure that the label is accurately adhered to the aluminum ingot.
[0082] In industrial robot control, movement along the tool refers to movement based on the tool coordinate system (usually based on the end of the tool). It usually involves the calculation of the position change (through x, y, z coordinates) and posture change (through rotation angles rx, ry, rz) of the tool end in three-dimensional space. The tool coordinate system has a specific transformation relative to the robot base coordinate system (or world coordinate system), which is usually represented by a homogeneous transformation matrix:
[0083]
[0084] Where R is the rotation matrix of 3′3 and t is the translation vector of 3′1. All movement along the tool is divided into two parts: rotation and translation. Rotation refers to the rotation of the tool end around its own coordinate system. Assume that the rotation of the tool coordinate system is expressed as Euler angles [rx, ry, rz]. It can be calculated by converting the Euler angles to a rotation matrix. If you need to rotate the tool coordinate system by an incremental angle [Drx, Dry, Drz], first convert it to the rotation matrix R D .
[0085] R D =R x (Drx)×R y (Dry)×R z (Drz)
[0086] The new rotation matrix R￠ is calculated as:
[0087] R￠=R×R D
[0088] Translation refers to the displacement of the tool end in the x, y, and z directions. Assume that the translation vector along the tool coordinate system is:
[0089] d=[dx,dy,dz] T
[0090] The new translation vector t￠ can be calculated by the following formula:
[0091] t￠=t+R×d
[0092] The calculation steps are to calculate the translation transformation and the rotation transformation, and finally update the transformation matrix T￠. The formula is:
[0093]
[0094] The method of the present invention supports various types of aluminum ingots, including aluminum ingots of different sizes, shapes and stacking methods, and adapts to different production requirements by adjusting the filtering algorithm parameters.
[0095] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. 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 identifying the position of aluminum ingot labeling based on 3D vision, characterized in that: The steps include: S1: Obtain point cloud data. The camera installed on the robot arm is aimed at the aluminum ingot to be labeled to take pictures and scan, and obtain the point cloud data of the aluminum ingot surface; S2: performing point cloud preprocessing on the acquired aluminum ingot surface point cloud data to filter out point cloud noise; S3: After the noise is filtered out, the point cloud is fitted into the shape of the aluminum ingot in a rectangular manner, and the labeling position is calculated based on the corner points of the fixed position of the fitted rectangle; S4: The labeling position calculated based on 3D vision is sent to the robot, and the robot labels the target aluminum ingot.
2. The method for identifying the position of aluminum ingot labeling based on 3D vision according to claim 1 is characterized in that: In the step S2, the point cloud preprocessing uses a filtering algorithm to perform noise reduction processing, and the filtering algorithm includes voxel grid sampling, side wall filtering, statistical denoising, and region segmentation.
3. The method for identifying the position of aluminum ingot labeling based on 3D vision according to claim 2 is characterized in that: The voxel grid sampling method is as follows: The point cloud is divided into three-dimensional grids, each grid is represented by a representative point, which is used to reduce the density of point cloud data and simplify the data. The formula is: Among them, p i is the point in the grid and N is the number of points in the grid.
4. The method for identifying the position of aluminum ingot labeling based on 3D vision according to claim 2 is characterized in that: The sidewall filtering method is as follows: Side wall filtering mainly uses normal vector analysis to calculate the normal vector of each point and determine whether it belongs to the side wall based on the direction of the normal vector. If the angle between the normal vector and the horizontal plane exceeds a certain threshold, it is considered to be a side wall point and filtered. The formula is as follows: Where n is the normal vector of the point and v is the reference direction vector.
5. The method for identifying the position of aluminum ingot labeling based on 3D vision according to claim 2 is characterized in that: The statistical denoising method is as follows: The average distance between each point and its neighboring points is calculated. If the distance exceeds a certain statistical threshold, the point is considered an outlier. The formula is: threshold=μ+ασ Among them, p is the point before w, p i are k neighboring points, μ is the average distance of all points, σ is the standard deviation, and α is the adjustment parameter.
6. The method for identifying the position of aluminum ingot labeling based on 3D vision according to claim 2 is characterized in that: The region segmentation method is as follows: Use Euclidean distance to measure the distance between two points in space. For two points p = (x1, y1, z1) and q = (x2, y2, z2) in a three-dimensional point cloud, the Euclidean distance is defined as: This formula is used to calculate the straight-line distance between points to determine their proximity in space. In the Euclidean distance segmentation method, the Euclidean clustering extraction algorithm is used.
7. The method for identifying the position of aluminum ingot labeling based on 3D vision according to claim 1 or 2, characterized in that: The rectangle fitting process of step S3 includes performing principal component analysis on the aluminum ingot point cloud, determining the main direction of the point cloud by calculating the covariance matrix of the point cloud data, and realizing the fitting of the minimum circumscribed rectangle, and the steps are: S3a: Calculate the point cloud centroid Among them, p i is the i-th point in the point cloud, N is the total number of points, and c is the centroid; S3b: Calculate the covariance matrix S3c: Calculate the eigenvalues and eigenvectors of the covariance matrix. The eigenvectors correspond to the main directions of the point cloud. Arrange the eigenvalues and select the eigenvectors corresponding to the two largest eigenvalues as the main directions. S3d: Rectangle fitting, projects the point cloud in two main directions and calculates the minimum bounding box of the projection points to obtain the four vertices of the rectangle. By adding a tool offset to one of the corner points of the rectangle, the labeling position of the aluminum ingot can be obtained by moving along the tool.