Robot automatic chamfering method and device based on laser vision

Abstract

The invention relates to a robot automatic chamfering method and device based on laser vision, and the method comprises the steps: scanning laser stripes on the surface of a welding seam, and obtaining a laser stripe image; image preprocessing is conducted on the laser stripe image, two-dimensional pixel point coordinates of the preprocessed laser stripe image are extracted, the two-dimensional pixel point coordinates are converted into three-dimensional coordinates under a robot base coordinate system, and point cloud data of the surface of the welding seam are obtained; segmenting the point cloud data to obtain an outer edge point set of the to-be-welded flat plate, and based on the outer edge point set, obtaining an actual polishing pose which should be adopted by the robot during chamfering polishing; the robot automatically changes to a safe pose which does not collide with the surrounding environment or the workpiece, a smooth transition path from the safe pose to an actual grinding pose is calculated, and a continuous movement track of the robot is obtained; and based on the continuous movement track, the robot completes chamfering and grinding operation. The method has the remarkable advantages in the aspects of improving the production efficiency, improving the intelligent degree, improving the safety and the like.

Description

Technical Field

[0001] The invention relates to the technical field of metal structural parts processing, and in particular to a robot automatic chamfering method and device based on laser vision. Background Art

[0002] In the field of precision manufacturing of high-end steel plate structural parts, the fine processing of the workpiece edge has become a crucial link for specific process requirements. Among them, achieving precise 45° chamfering of the edge is not only related to the aesthetics of the product, but also directly affects the performance stability of the product and the smoothness of subsequent assembly. However, in current production practice, this key link generally relies on traditional manual grinding and human-machine collaboration mode, and its inherent challenges and limitations are becoming increasingly prominent.

[0003] Specifically, although manual chamfering with a handheld angle grinder can accomplish the task to a certain extent with the craftsman's exquisite skills, this method is highly dependent on the operator's experience and physical condition, resulting in low work efficiency and difficulty in ensuring consistency in the chamfering effect between finished products, which not only increases the difficulty of quality control but also affects the overall quality of the product. In addition, long-term exposure to high-noise and dusty working environments poses a potential threat to workers' health, and the high labor costs have also become a burden that companies cannot ignore.

[0004] Although the human-machine collaborative polishing mode has reduced the labor burden to a certain extent, it has not yet completely gotten rid of the dependence on manual fine assembly. This not only limits the further improvement of the degree of automation, but also pushes up production costs due to the need for additional human resources investment. At the same time, it reduces the flexibility and response speed of the overall production line, making it difficult to meet the urgent needs of modern manufacturing for high efficiency, precision and low cost.

[0005] Therefore, exploring and introducing more advanced and intelligent automated chamfering processing solutions has become an urgent need and inevitable trend for the transformation and upgrading of the steel plate structural parts manufacturing industry. Summary of the invention

[0006] To this end, the technical problem to be solved by the present invention is to overcome the problems of low efficiency of chamfering operations and low degree of automation in the human-machine collaborative mode in the prior art. The present invention aims to propose an innovative technical solution, which will be committed to improving the level of automation of chamfering processing by introducing intelligent equipment, optimizing process flow and integrating advanced control technology to achieve more efficient, more precise and more stable chamfering operations.

[0007] In order to solve the above technical problems, the present invention provides a robot automatic chamfering method based on laser vision, comprising the following steps:

[0008] S1: Projecting laser onto the weld surface of the workpiece to be identified, scanning laser stripes on the weld surface, and obtaining a laser stripe image;

[0009] S2: performing image preprocessing on the laser stripe image to obtain a preprocessed laser stripe image;

[0010] S3: extracting two-dimensional pixel point coordinates based on the preprocessed laser stripe image, converting the two-dimensional pixel point coordinates into three-dimensional coordinates in a robot base coordinate system, and obtaining point cloud data of the weld surface;

[0011] S4: segmenting the point cloud data to obtain an outer edge point set of the plate to be welded, and based on the outer edge point set, obtaining an actual grinding posture that the robot should take when performing chamfer grinding;

[0012] S5: The robot automatically changes to a safe posture that does not collide with the surrounding environment or the workpiece, calculates a smooth transition path from the safe posture to the actual grinding posture, and obtains a continuous motion trajectory of the robot;

[0013] S6: Based on the continuous motion trajectory, the robot completes the chamfering and grinding operation.

[0014] In one embodiment of the present invention, the method for obtaining a laser stripe image includes: continuously acquiring the laser stripe image at a constant scanning rate in the scanning direction; or acquiring the laser stripe image when reaching the scanning point by setting interval scanning points in the scanning direction.

[0015] In one embodiment of the present invention, the method for obtaining a preprocessed laser stripe image includes: denoising, enhancing and edge detecting the laser stripe image to obtain the preprocessed laser stripe image.

[0016] In one embodiment of the present invention, the method for extracting two-dimensional pixel point coordinates is: first, threshold segmentation is performed on the preprocessed laser stripe image, the weld area in the preprocessed laser stripe image is separated from the background, and its two-dimensional pixel point coordinates are extracted.

[0017] In one embodiment of the present invention, the method for obtaining the outer edge point set of the plate to be welded is:

[0018] S41: Calculate the statistical characteristic value between each point in the point cloud data and its neighboring points, compare the statistical characteristic value with a set noise point statistical threshold, remove the noise points in the point cloud data, and obtain denoised point cloud data;

[0019] S42: performing downsampling processing on the denoised point cloud data to obtain downsampled point cloud data;

[0020] S43: dividing the point cloud data into multiple clusters with similar features according to the similarity between the pixels in the downsampled point cloud data;

[0021] S44: clustering the plurality of clusters of point cloud data having similar features to obtain cluster segmented point cloud data;

[0022] S45: Based on the point cloud data after clustering and segmentation, a K-dimensional tree structure is constructed. For each point in the point cloud, the K-dimensional tree structure is used to query its N nearest points. If the number of neighboring points of a point is less than a preset threshold M, the point is recorded as an outlier point, and the outlier point is removed from the point cloud data after clustering and segmentation to obtain updated point cloud data.

[0023] S46: Based on the updated point cloud data, the normal and curvature of each point are calculated to identify the plane area, and the outer edge point set of the plane area is obtained, that is, the outer edge point set of the plate to be welded;

[0024] The statistical characteristic value includes the average distance or standard deviation between each point in the point cloud data and its neighboring points.

[0025] In one embodiment of the present invention, the method for obtaining downsampled point cloud data is:

[0026] S421: Divide the denoised point cloud data into a plurality of three-dimensional voxel grids, set a downsampling target value, and set the size of each voxel grid according to the point cloud data density and the downsampling target value;

[0027] S422: traverse all point cloud data, assign each point to a corresponding voxel grid according to its spatial coordinates, and assign a point located on the boundary of multiple voxel grids to a grid with the shortest average distance;

[0028] S423: Calculate the centroid of all points in each voxel grid, and use the centroid as a representative point to represent all points in the voxel grid;

[0029] S424: Collect representative points in each voxel grid to obtain downsampled point cloud data.

[0030] In one embodiment of the present invention, a method for obtaining multiple clusters of point cloud data having similar features is as follows:

[0031] S431: Based on the downsampled point cloud data, arbitrarily select one or more pixel points as initial seed points;

[0032] S432: The initial seed point is taken as a region and added to a region set. In each iteration, a region is selected from the region set as the current region, and the neighboring pixels of the current region are traversed. If the neighboring pixels meet the similarity condition with the current region, they are added to the current region and marked as visited.

[0033] S433: Repeat step S432 until all pixels have been visited or no new pixels can be added to any area, thereby obtaining multiple clusters of point cloud data with similar features.

[0034] In one embodiment of the present invention, the method for obtaining the outer edge point set of the plane area is:

[0035] S461: Selecting the leftmost, rightmost, topmost, bottommost, frontmost and last points from the point set of the plane area as vertices of the initial convex hull;

[0036] S462: Divide the point set into multiple parts through the edge of the initial convex hull, each part contains points located outside the current convex hull surface;

[0037] S463: For each segmented part, calculate the maximum distance point between the current face and the external point set, select the point with the largest distance to add to the convex hull, and update the current face;

[0038] S464: Repeat steps S461 to S463 until all points are contained in the convex hull, and obtain a minimum convex hull consisting of a series of vertices arranged in order, and obtain the outer edge point set of the plane area based on the minimum convex hull.

[0039] In one embodiment of the present invention, the actual grinding posture includes the posture and position of the grinding tool, and the acquisition method thereof is:

[0040] Performing interpolation processing on the outer edge point set to obtain initial boundary point coordinates, and fitting the initial boundary point coordinates to obtain a continuous curve-shaped boundary representation;

[0041] Based on the boundary representation, obtaining a plane normal vector and an edge line direction vector;

[0042] According to the tool size, the grinding surface of the tool is made perpendicular to the plane normal vector, and the moving direction or main axis direction of the tool is consistent with or at a certain angle to the edge line direction vector to determine the posture and position of the grinding tool.

[0043] The present invention also provides a robot automatic chamfering device based on laser vision, comprising:

[0044] A workpiece chamfering actuator, wherein the workpiece chamfering actuator is mounted at the end of the robot;

[0045] A laser vision sensor, wherein the laser vision sensor is arranged on the workpiece chamfering actuator;

[0046] And a control system, the control system is connected to the workpiece chamfering actuator and the laser vision sensor, and during the chamfering and grinding process, the control system executes the instructions of the laser vision-based robot automatic chamfering method.

[0047] The above technical solution of the present invention has the following advantages compared with the prior art:

[0048] 1. High degree of automation: The laser stripe image on the weld surface is collected in real time through the laser vision sensor, and the image processing and point cloud data extraction are automatically performed. Finally, the robot is controlled to automatically complete the chamfering and grinding operation, which greatly improves the level of production automation and reduces manual intervention.

[0049] 2. High processing efficiency: Compared with traditional manual grinding and human-machine collaborative grinding, the automated method of the present invention can perform chamfering processing continuously and quickly, which significantly improves the operating efficiency and shortens the production cycle.

[0050] 3. Quality assurance: The use of laser vision sensors combined with image processing technology can accurately extract point cloud data on the weld surface, and calculate accurate grinding postures based on these data to ensure consistent grinding effects each time, thereby improving product quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to make the content of the present invention more clearly understood, the present invention is further described in detail below according to specific embodiments of the present invention in conjunction with the accompanying drawings, wherein

[0052] Figure 1 is a flow chart of a robot automatic chamfering method based on laser vision provided in an embodiment of the present invention;

[0053] Figure 2 is a flow chart of a method for obtaining an outer edge point set of a flat plate to be welded provided in an embodiment of the present invention;

[0054] Figure 3 is a flow chart of a method for obtaining downsampled point cloud data provided in an embodiment of the present invention;

[0055] Figure 4 is a flow chart of a method for obtaining multiple clusters of point cloud data having similar features provided in an embodiment of the present invention;

[0056] Figure 5 is a flow chart of a method for obtaining an outer edge point set of the planar region provided in an embodiment of the present invention;

[0057] Figure 6 is a schematic structural diagram of a robot automatic chamfering device based on laser vision provided in an embodiment of the present invention;

[0058] Explanation of the reference numerals in the specification: 10. Workpiece chamfering actuator; 20. Laser vision sensor; 30. Control system. DETAILED DESCRIPTION

[0059] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments so that those skilled in the art can better understand the present invention and implement it, but the embodiments are not intended to limit the present invention.

[0060] Embodiment 1

[0061] Reference Figures 1 to 5 As shown, the present invention provides a robot automatic chamfering method based on laser vision, comprising the following steps:

[0062] S1: Projecting laser onto the weld surface of the workpiece to be identified through a laser vision sensor, the camera in the laser sensor simultaneously starts scanning the laser stripe signal reflected back from the weld surface to obtain a laser stripe image containing weld surface morphology information;

[0063] S2: performing image denoising, enhancement and edge detection preprocessing operations on the laser stripe image through the built-in image processing module of the laser vision sensor to obtain a preprocessed laser stripe image;

[0064] S3: extracting two-dimensional pixel point coordinates based on the preprocessed laser stripe image, and converting the two-dimensional pixel point coordinates into three-dimensional coordinates in the robot base coordinate system through laser plane conversion and hand-eye conversion to obtain point cloud data of the weld surface;

[0065] S4: segmenting the point cloud data to obtain an outer edge point set of the plate to be welded, and based on the outer edge point set, obtaining an actual grinding posture that the robot should take when performing chamfer grinding;

[0066] S5: The robot automatically changes to a safe posture that does not collide with the surrounding environment or the workpiece, calculates a smooth transition path from the safe posture to the actual grinding posture, and obtains a continuous motion trajectory of the robot;

[0067] S6: Based on the continuous motion trajectory, the robot completes the chamfering and grinding operation.

[0068] In this embodiment, the method for obtaining the laser stripe image includes: continuously collecting the laser stripe image at a constant scanning rate in the scanning direction; or collecting the laser stripe image when reaching the scanning point by setting interval scanning points in the scanning direction.

[0069] Specifically, in step S3, the method for extracting the two-dimensional pixel coordinates is: firstly, threshold segmentation is performed on the preprocessed laser stripe image, the weld area in the preprocessed laser stripe image is separated from the background, and its two-dimensional pixel coordinates are extracted.

[0070] Reference Figure 2 As shown, in step S4, in order to effectively remove the interference noise points caused by the laser reflection in the original point cloud data and the influence of the laser stripes in the non-weld area, the point cloud data is segmented to obtain the outer edge point set of the plate to be welded, and the method steps are as follows:

[0071] S41: Calculate the statistical characteristic value between each point in the point cloud data and its neighboring points, compare the statistical characteristic value with a set noise point statistical threshold, record the points greater than the noise point statistical threshold as noise points, remove the noise points in the point cloud data, and obtain denoised point cloud data;

[0072] S42: performing downsampling processing on the denoised point cloud data to obtain downsampled point cloud data;

[0073] S43: dividing the point cloud data into multiple clusters with similar features according to the similarity between the pixels in the downsampled point cloud data;

[0074] S44: performing cluster segmentation on the multiple clusters of point cloud data having similar features, calculating the Euclidean distance between each data point, and obtaining point cloud data after cluster segmentation;

[0075] S45: Based on the point cloud data after clustering and segmentation, a K-dimensional tree structure is constructed. For each point in the point cloud, the K-dimensional tree structure is used to query its N nearest points. If the number of neighboring points of a point is less than a preset threshold M, the point is recorded as an outlier point, and the outlier point is removed from the point cloud data after clustering and segmentation to obtain updated point cloud data.

[0076] S46: Based on the updated point cloud data, the normal and curvature of each point are calculated, and combined with the information restriction in the Z direction, a plane area with specific plane characteristics is identified and extracted, that is, the plate where the weld is located, and an outer edge point set of the plane area is obtained from the plane area, that is, the outer edge point set of the plate to be welded;

[0077] The statistical characteristic value includes the average distance or standard deviation between each point in the point cloud data and its neighboring points.

[0078] Further, in step S42, the denoised point cloud data is downsampled by a voxel downsampling filtering method, which can reduce the point cloud density and the amount of calculation to obtain downsampled point cloud data, such as Figure 3 As shown, the specific method steps are:

[0079] S421: Divide the denoised point cloud data into a plurality of three-dimensional voxel grids, set a downsampling target value, and set the size of each voxel grid according to the point cloud data density and the downsampling target value;

[0080] S422: traverse all point cloud data, assign each point to a corresponding voxel grid according to its spatial coordinates, and assign a point located on the boundary of multiple voxel grids to a grid with the shortest average distance;

[0081] S423: Calculate the centroid of all points in each voxel grid, and use the centroid as a representative point to represent all points in the voxel grid;

[0082] S424: Collect representative points in each voxel grid to obtain downsampled point cloud data.

[0083] Furthermore, if Figure 4 As shown, in step S43, according to the similarity between the pixels in the downsampled point cloud data, the point cloud data is divided into multiple clusters with similar features, and the specific method steps are as follows:

[0084] S431: Based on the downsampled point cloud data, arbitrarily select one or more pixel points as initial seed points;

[0085] S432: The initial seed point is taken as a region and added to a region set. In each iteration, a region is selected from the region set as the current region, and the neighboring pixels of the current region are traversed. If the neighboring pixels meet the similarity condition with the current region, they are added to the current region and marked as visited.

[0086] S433: Repeat step S432 until all pixels have been visited or no new pixels can be added to any area, thereby obtaining multiple clusters of point cloud data with similar features.

[0087] In order to extract the edge position of the plate surface to be polished, the edge of the plate to be polished is obtained by using the convex hull calculation method, that is, the outer edge point set of the plane area. In three-dimensional space, the convex hull is the smallest convex polyhedron that encloses a given point set. Each face of the convex hull is a plane polygon. The basic idea of ​​calculating the convex hull is to find the smallest polyhedron that can enclose all points, each face of which is composed of points in the point set.

[0088] Furthermore, if Figure 5 As shown, in step S46, the steps of the method for obtaining the outer edge points of the plane area are:

[0089] Selecting a starting point: S461: Based on the divide-and-conquer method of Quickhull quick sort, selecting the leftmost, rightmost, topmost, bottommost, frontmost and last points from the point set of the plane area as the vertices of the initial convex hull;

[0090] Perform point set segmentation: S462: segment the point set into multiple parts through the edge of the initial convex hull, each part including points located outside the current convex hull surface;

[0091] Recursively construct the minimum convex hull:

[0092] S463: For each segmented part, calculate the maximum distance point between the current face and the external point set, select the point with the largest distance to add to the convex hull, and update the current face;

[0093] S464: Repeat steps S461 to S463 until all points are contained in the convex hull, and obtain a minimum convex hull consisting of a series of vertices arranged in order, and obtain the outer edge point set of the plane area based on the minimum convex hull.

[0094] The outer edge point set of the plane area obtained by the above method is segmented, fitted and interpolated, and the chamfered edges are numbered to facilitate the setting and adjustment of the subsequent chamfering operation sequence.

[0095] Furthermore, based on the outer edge point set, the actual grinding posture that the robot should take when performing chamfer grinding is obtained, and the actual grinding posture includes the posture and position of the grinding tool, and the acquisition method is:

[0096] Performing interpolation processing on the outer edge point set to obtain initial boundary point coordinates, and fitting the initial boundary point coordinates to obtain a continuous curve-shaped boundary representation;

[0097] Based on the boundary representation, obtaining a plane normal vector and an edge line direction vector;

[0098] According to the tool size, the grinding surface of the tool is made perpendicular to the plane normal vector, and the moving direction or main axis direction of the tool is consistent with or at a certain angle to the edge line direction vector to determine the posture and position of the grinding tool.

[0099] According to the actual grinding posture, the robot posture is automatically changed to run to the robot safe posture of the "grinding stage". Then, the average posture is calculated according to the output grinding tool posture, and the transformation from the safe posture to this posture is performed to avoid the interference problem caused by the posture transformation during the grinding process. Finally, the six-element parameters calculated above are sent to the control end of the robot end in the form of a point sequence, and the control end is used to control the robot end to perform the chamfering grinding process.

[0100] Embodiment 2

[0101] Reference Figure 6 As shown, the present invention also provides a robot automatic chamfering device based on laser vision, which includes the following components:

[0102] A workpiece chamfering actuator 10, wherein the workpiece chamfering actuator 10 is mounted at the end of the robot;

[0103] A laser vision sensor 20, which is disposed on the workpiece chamfering actuator 10 and is used to identify and locate the workpiece to be chamfered;

[0104] And a control system 30, the control system 30 is connected to the workpiece chamfering actuator 10 and the laser vision sensor 20, and during the chamfering grinding process, the control system 30 executes the instructions of the robot automatic chamfering method based on laser vision described in Example 1.

[0105] During the automatic chamfering process, the control system 30 controls the robot arm to carry the workpiece chamfering actuator 10 and the laser vision sensor 20 to move together, first changing the robot posture to restore the robot posture in the "scanning phase". The robot arm will be lifted to a preset safe scanning position and move forward according to the scanning path.

[0106] At the same time, the control system 30 sends an operation instruction to the laser vision sensor 20, and the laser vision sensor 20 starts the laser emitter, projects the laser line onto the surface of the workpiece to be identified, captures the laser stripes through the grayscale camera of the laser vision sensor 20, and saves the laser stripe image during the laser scanning process and sends it to the control system 30.

[0107] The control system 30 processes the laser stripe image data, and after laser three-dimensional reconstruction and adaptive chamfering posture generation, sends the target point position and posture to the robot end, and the robot and its workpiece chamfering actuator 10 complete the actual chamfering and grinding process.

[0108] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0109] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0110] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0111] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0112] Obviously, the above embodiments are merely examples for clear explanation and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived from these are still within the protection scope of the invention.

Claims

1. A robot automatic chamfering method based on laser vision, characterized in that: The following steps are involved: S1: Projecting laser onto the weld surface of the workpiece to be identified, scanning laser stripes on the weld surface, and obtaining a laser stripe image; S2: performing image preprocessing on the laser stripe image to obtain a preprocessed laser stripe image; S3: extracting two-dimensional pixel point coordinates based on the preprocessed laser stripe image, converting the two-dimensional pixel point coordinates into three-dimensional coordinates in a robot base coordinate system, and obtaining point cloud data of the weld surface; S4: segmenting the point cloud data to obtain an outer edge point set of the plate to be welded, and based on the outer edge point set, obtaining an actual grinding posture that the robot should take when performing chamfer grinding; S5: The robot automatically changes to a safe posture that does not collide with the surrounding environment or the workpiece, calculates a smooth transition path from the safe posture to the actual grinding posture, and obtains a continuous motion trajectory of the robot; S6: Based on the continuous motion trajectory, the robot completes the chamfering and grinding operation.

2. The robot automatic chamfering method based on laser vision according to claim 1 is characterized in that: The method for obtaining the laser stripe image includes: continuously collecting the laser stripe image at a constant scanning rate in the scanning direction; or collecting the laser stripe image when reaching the scanning point by setting interval scanning points in the scanning direction.

3. The robot automatic chamfering method based on laser vision according to claim 1 is characterized in that: The method for obtaining the preprocessed laser stripe image comprises: denoising, enhancing and edge detecting the laser stripe image to obtain the preprocessed laser stripe image.

4. The robot automatic chamfering method based on laser vision according to claim 1 is characterized in that: The method for extracting two-dimensional pixel point coordinates is: firstly, threshold segmentation is performed on the preprocessed laser stripe image, the weld area in the preprocessed laser stripe image is separated from the background, and its two-dimensional pixel point coordinates are extracted.

5. The robot automatic chamfering method based on laser vision according to claim 1 is characterized in that: The method to obtain the outer edge point set of the plate to be welded is: S41: Calculate the statistical characteristic value between each point in the point cloud data and its neighboring points, compare the statistical characteristic value with a set noise point statistical threshold, remove the noise points in the point cloud data, and obtain denoised point cloud data; S42: performing downsampling processing on the denoised point cloud data to obtain downsampled point cloud data; S43: dividing the point cloud data into multiple clusters with similar features according to the similarity between the pixels in the downsampled point cloud data; S44: clustering the plurality of clusters of point cloud data having similar features to obtain cluster segmented point cloud data; S45: Based on the point cloud data after clustering and segmentation, a K-dimensional tree structure is constructed. For each point in the point cloud, the K-dimensional tree structure is used to query its N nearest points. If the number of neighboring points of a point is less than a preset threshold M, the point is recorded as an outlier point, and the outlier point is removed from the point cloud data after clustering and segmentation to obtain updated point cloud data. S46: Based on the updated point cloud data, the normal and curvature of each point are calculated to identify the plane area, and the outer edge point set of the plane area is obtained, that is, the outer edge point set of the plate to be welded; The statistical characteristic value includes the average distance or standard deviation between each point in the point cloud data and its neighboring points.

6. The robot automatic chamfering method based on laser vision according to claim 5 is characterized in that: The method to obtain the downsampled point cloud data is: S421: Divide the denoised point cloud data into a plurality of three-dimensional voxel grids, set a downsampling target value, and set the size of each voxel grid according to the point cloud data density and the downsampling target value; S422: traverse all point cloud data, assign each point to a corresponding voxel grid according to its spatial coordinates, and assign a point located on the boundary of multiple voxel grids to a grid with the shortest average distance; S423: Calculate the centroid of all points in each voxel grid, and use the centroid as a representative point to represent all points in the voxel grid; S424: Collect representative points in each voxel grid to obtain downsampled point cloud data.

7. The robot automatic chamfering method based on laser vision according to claim 5 is characterized in that: The method to obtain multiple clusters of point cloud data with similar features is: S431: Based on the downsampled point cloud data, arbitrarily select one or more pixel points as initial seed points; S432: The initial seed point is taken as a region and added to a region set. In each iteration, a region is selected from the region set as the current region, and the neighboring pixels of the current region are traversed. If the neighboring pixels meet the similarity condition with the current region, they are added to the current region and marked as visited. S433: Repeat step S432 until all pixels have been visited or no new pixels can be added to any area, thereby obtaining multiple clusters of point cloud data with similar features.

8. The robot automatic chamfering method based on laser vision according to claim 5 is characterized in that: The method for obtaining the outer edge point set of the plane area is: S461: Selecting the leftmost, rightmost, topmost, bottommost, frontmost and last points from the point set of the plane area as vertices of the initial convex hull; S462: Divide the point set into multiple parts through the edge of the initial convex hull, each part contains points located outside the current convex hull surface; S463: For each segmented part, calculate the maximum distance point between the current face and the external point set, select the point with the largest distance to add to the convex hull, and update the current face; S464: Repeat steps S461 to S463 until all points are contained in the convex hull, and obtain a minimum convex hull consisting of a series of vertices arranged in order, and obtain the outer edge point set of the plane area based on the minimum convex hull.

9. The robot automatic chamfering method based on laser vision according to claim 1 is characterized in that: The actual grinding posture includes the posture and position of the grinding tool, and the method for obtaining the posture and position of the grinding tool is as follows: Performing interpolation processing on the outer edge point set to obtain initial boundary point coordinates, and fitting the initial boundary point coordinates to obtain a continuous curve-shaped boundary representation; Based on the boundary representation, obtaining a plane normal vector and an edge line direction vector; According to the tool size, the grinding surface of the tool is made perpendicular to the plane normal vector, and the moving direction or main axis direction of the tool is consistent with or at a certain angle to the edge line direction vector to determine the posture and position of the grinding tool.

10. A robot automatic chamfering device based on laser vision, characterized in that: include: A workpiece chamfering actuator, wherein the workpiece chamfering actuator is mounted at the end of the robot; A laser vision sensor, wherein the laser vision sensor is arranged on the workpiece chamfering actuator; And a control system, the control system is connected to the workpiece chamfering actuator and the laser vision sensor, and during the chamfering grinding process, the control system executes the instructions of the robot automatic chamfering method based on laser vision as described in any one of claims 1 to 9.

Images