Turbine blade wear-resistant layer welding path planning method based on point cloud processing

Abstract

The invention provides a turbine blade wear-resistant layer welding path planning method based on point cloud processing. The method comprises the steps that firstly, a structured light camera and a welding robot tail end coordinate system and a conversion matrix of the welding robot tail end and a world coordinate system are determined through calibration; scanning the turbine blade by using a structured light camera to obtain original three-dimensional point cloud data of the turbine blade, and recording coordinates of a welding gun under a tail end coordinate system of a welding robot; performing point cloud filtering, Euclidean clustering and plane segmentation on the original three-dimensional point cloud data to obtain point cloud data of a to-be-welded surface; the point cloud of the surface to be welded is cut through a principal component analysis algorithm to obtain a preliminary path, a complete path can be obtained through straight line fitting and path logic sorting processing, and then coordinates of a welding gun in a coordinate system at the tail end of the welding robot and the complete path are substituted into a coordinate transformation matrix; coordinates of all the path points in the world coordinate system can be obtained, so that path planning is completed. According to the method, automatic path planning can be realized, and production efficiency is improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of turbine blade welding of an aero-engine, and particularly relates to a turbine blade wear-resistant layer welding path planning method based on point cloud processing. BACKGROUND

[0002] Turbine blades are key parts of an aero-engine. During operation of the engine, extremely high temperature and pressure are generated, and high-speed rotation generates intense vibration and strong centrifugal force. Based on these factors, the service environment of the engine turbine blade is very harsh. Therefore, the performance and precision of the engine turbine blade manufacturing have very high requirements. In the currently adopted turbine blade design, a crown structure is often used to reduce vibration, and the adjacent two turbine blades are meshed with each other to form a blade ring through the crown protruding tip. By using welding method to preinstall wear-resistant layer at the meshing position, the meshing gap can be reduced, and the influence of wear and vibration can be reduced, thereby improving the service life of the turbine blade.

[0003] Path planning is an important concept in welding. For the process of surfacing, the possibility of directly forming a single straight line path using a welding gun is very low, so the welding path needs to be designed according to the characteristics of the surfacing area. However, in the actual production, the path planning of most large simple planar rectangular structures is carried out, and the planning of complex shapes such as engine turbine blades is less, the degree of automation is low, and in actual production, workers usually still rely on manual surfacing. The wear-resistant layer area of the turbine blade is very small, only 3-5 mm, and it is difficult to meet the actual production needs by relying on the naked eye to process. Even if some factories can rely on teaching robots to weld, the turbine blades must be placed according to fixed positions and angles, the degree of intelligence is low, and it is difficult for the welding gun to change according to the cross-sectional shape of the part, so the turbine blade surfacing quality is still not ideal. In general, the current intelligent degree of turbine blade wear-resistant layer structure welding is low, and the precision is not ideal.

[0004] In the prior art, some turbine blade layer welding path planning methods with high degree of intelligence are also disclosed, for example: the Chinese patent application with the patent application number CN202210444549.4 and the invention name of laser cladding path planning method and system for the top end of an aero-engine blade, discloses the following content: first, a binary image of the top end section of the blade is obtained and inverted, and the skeleton of the inverted image is extracted; the inverted image and the skeleton are calculated by the contour offset method to generate a preliminary cladding path; the preliminary path is cut and repaired to optimize the path; finally, the repaired path is processed by the contour line method to obtain the final laser cladding path planning diagram, so as to achieve more uniform cladding effect.

[0005] The above scheme has the following disadvantages: (1) The locality of path planning: its path planning is mainly based on the binary image and skeleton extraction of the turbine blade tip, which may not fully consider the overall three-dimensional geometric characteristics of the turbine blade; for complex damage or non-uniform damage, the path adaptability may be insufficient.

[0006] (2) Multiple and complex operation steps: the above scheme involves multiple expansion, boundary extraction, pixel point comparison and interpolation calculation, which may result in high computational complexity, especially when processing high-resolution images or complex turbine blade shapes, the calculation efficiency may be low. SUMMARY

[0007] In order to solve the problems existing in the prior art, the application provides a turbine blade wear-resistant layer welding path planning method based on point cloud processing.

[0008] In order to achieve the above purpose, the application provides a turbine blade wear-resistant layer welding path planning method based on point cloud processing, comprising the following steps: Step 1, fix the structured light camera on the welding gun, calibrate the structured light camera, obtain the conversion matrix between the structured light camera coordinate system and the welding robot end coordinate system; and obtain the conversion matrix between the welding robot end coordinate system and the world coordinate system; Step 2, scan a turbine blade using the structured light camera to obtain the original three-dimensional point cloud data of the turbine blade, and record the coordinates of the welding gun in the welding robot end coordinate system; Step 3, pre-process the original three-dimensional point cloud data obtained in step 2 to obtain a point cloud region containing point cloud data corresponding to the welding surface; Step 4, the welding surface is obtained, and the point cloud data corresponding to the welding surface is obtained; Step 5, preliminarily screen the path point cloud data in the point cloud data corresponding to the welding surface, and screen out several preliminary paths from the welding surface; Step 6, process the point cloud data corresponding to each preliminary path obtained in step 5, and for each preliminary path, a linear point cloud will be obtained, and m points are selected from the linear point cloud, denoted as a group of path points, m is a positive integer, m≥2; Step 7, arrange the corresponding several groups of path points on the welding surface in the following way: arrange the serial numbers of the points in the path points in the odd number of groups in the forward direction; arrange the serial numbers of the points in the path points in the even number of groups in the reverse direction; finally form a complete path with the corresponding several groups of path points on the welding surface. Step 8, the coordinates of the welding torch recorded in the coordinate system of the welding robot end in step 2 and the coordinates of the structured light camera coordinate system corresponding to the complete path obtained in step 7 are substituted into the coordinate conversion matrix obtained in step 1, the coordinates of the path points in the structured light camera coordinate system are converted into the coordinates in the world coordinate system, and the actual coordinate values of all path points are obtained, and thus the path planning is completed.

[0009] In some embodiments, the preprocessing process involved in step 3 is as follows: Step 301, using a pass-through filter to remove invalid points in the original point cloud data and remove background point cloud data; Step 302, using voxel filtering to reconstruct the density of the point cloud obtained in step 301; Step 303, using the K-Means clustering method to segment the density-reconstructed point cloud into several point cloud regions, and screening out the point cloud region containing the welding surface.

[0010] In some embodiments, the processing process of step 4 is as follows: using the Ransac plane segmentation algorithm, the outer surface of the screened point cloud region containing the welding surface is segmented into several independent surfaces, and the useless point cloud data is removed, and the point cloud data corresponding to the welding surface is retained.

[0011] In some embodiments, the processing process of the welding surface in step 5 is as follows: Step 501, using the principal component analysis algorithm to centralize the point cloud data corresponding to the welding surface; Step 502, calculating the covariance matrix of the point cloud data corresponding to the welding surface, and calculating the eigenvalues and eigenvectors of the covariance matrix, the eigenvector corresponding to the maximum eigenvalue is the long side direction of the welding surface, the eigenvector corresponding to the second largest eigenvalue is the short side direction of the welding surface, and the eigenvector corresponding to the minimum eigenvalue is the normal vector direction of the welding surface. The welding surface is regarded as a rectangle, and the length of the long side and the length of the short side of the welding surface are obtained by using the dot product formula, and the height of the minimum bounding box corresponding to the point cloud of the welding surface is obtained according to the minimum bounding box. Step 503, using the three directions and three sizes obtained in step 502, a cutting surface perpendicular to the welding surface is generated on the welding surface, the intersection line of the cutting surface and the welding surface extends along the short side direction of the welding surface, and the distance between the intersection line and the short side of the left end of the welding surface is a first preset distance. Step 504, using the dot product formula, the point cloud data in the point cloud data corresponding to the welding surface within a second preset distance range from the current cutting surface is formed into a preliminary path; Step 505, judge whether the remaining length of the welding surface along the long edge direction after being cut by the current cutting surface is greater than the welding bead width, if yes, jump to step 506, if not, end; Step 506, generate a next cutting surface perpendicular to the welding surface along the long edge direction of the welding surface at a distance of one welding bead width from the current cutting surface, the adjacent two cutting surfaces are parallel to each other; using the dot product formula, the point cloud data in the second preset distance range from the new generated cutting surface corresponding to the point cloud data of the welding surface is formed into a preliminary path; Step 507, repeat step 505 and step 506 until the remaining length of the welding surface along the long edge direction after being cut by the last cutting surface is less than the welding bead width, thus obtaining all preliminary paths corresponding to the welding surface.

[0012] In some embodiments, in step 6, the point cloud data corresponding to one of the preliminary paths obtained in step 5 is processed to obtain a linear point cloud, and the processing process is as follows: the Ransac straight line segmentation algorithm is used to fit the point cloud data corresponding to the preliminary path to obtain a fitting path, at this time, the fitting path is in the form of a three-dimensional model, and the corresponding linear point cloud is obtained after the three-dimensional model corresponding to the fitting path is point cloudized.

[0013] The turbine blade wear-resistant layer welding path planning method based on point cloud processing has the following advantages: 1) The turbine blade wear-resistant layer welding process can realize automatic path planning, which can greatly improve the production efficiency; 2) The turbine blade wear-resistant layer welding process can improve the machining precision, and the error is within 0.05mm; 3) Reduce the placement requirements of the turbine blade in the production process, realize intelligentization and improve the production efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0014] Figure 1 A flow chart of the turbine blade wear-resistant layer welding path planning method based on point cloud processing in the embodiment is shown.

[0015] Figure 2 A schematic diagram of the original three-dimensional point cloud of the turbine blade in the embodiment is shown.

[0016] Figure 3 A schematic diagram of the point cloud region containing the welding surface screened out in the embodiment is shown.

[0017] Figure 4 A schematic diagram of the removed useless point cloud in the embodiment is shown.

[0018] Figure 5 A schematic diagram of a welding surface corresponding to a turbine blade in an embodiment is shown.

[0019] Figure 6 A schematic diagram of a cutting surface perpendicular to the welding surface in an embodiment is shown.

[0020] Figure 7 A schematic diagram of all the preliminary paths corresponding to the welding surface in an embodiment is shown.

[0021] Figure 8 A schematic diagram of a fitted path obtained by fitting the point cloud data corresponding to the preliminary paths in an embodiment is shown.

[0022] Figure 9 A schematic diagram of the arrangement of two adjacent groups of path points in an embodiment is shown.

[0023] Figure 10 A schematic diagram of a complete path formed on the welding surface in an embodiment is shown.

[0024] Reference signs: 10 - point cloud region containing a welding surface, 20 - welding surface, 30 - cutting surface, 301 - first corner point, 302 - second corner point, 303 - third corner point, 304 - fourth corner point, 40 - preliminary path, 50 - fitted path. DETAILED DESCRIPTION

[0025] The specific embodiments of the present application will be further described in conjunction with the accompanying drawings.

[0026] In the description of the present application, it should be understood that the terms "first", "second" and the like are used to distinguish similar objects, and are not used to describe or indicate specific order or sequence, and the terms "upper", "lower", "front", "back", "left", "right", "top", "bottom", "inner", "outer" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only used to facilitate the description of the present application and simplify the description, and therefore cannot be understood as indicating or implying that the device or element referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation on the present application.

[0027] Three-dimensional point cloud technology is actually to use a scanner to scan and abstract an object into points in a three-dimensional space, and to represent the object by these points in the three-dimensional space. Compared with images or models, three-dimensional point clouds are more suitable for abstraction into data, and are more convenient, rapid and accurate to process in a computer, and have a higher degree of automation in actual industrial production, which can greatly improve production efficiency and quality.

[0028] Based on this, the turbine blade wear-resistant layer welding path planning method based on point cloud processing involved in the present application includes the following steps, as shown in Figure 1 Step 1, fix the structured light camera on the welding gun, calibrate the structured light camera, obtain the conversion matrix between the structured light camera coordinate system and the welding robot end coordinate system (TCP), and obtain the conversion matrix between the welding robot end coordinate system and the world coordinate system.

[0029] Step 2, scan a turbine blade using the structured light camera to obtain the original three-dimensional point cloud data of the turbine blade, as shown in Figure 2 , while recording the TCP coordinates of the welding gun.

[0030] Step 3, pre-process the original three-dimensional point cloud data obtained in step 2 to obtain a point cloud region containing point cloud data corresponding to the welding surface 20.

[0031] Specifically, the pre-processing process involved in step 3 is as follows: Step 301, use straight-through filtering to remove invalid points in the original point cloud data and remove background point cloud data; Step 302, use voxel filtering to reconstruct the density of the point cloud obtained in step 301; Step 303, use the K-Means clustering method to segment the density-reconstructed point cloud into several point cloud regions, and select the point cloud region 10 containing the welding surface, as shown in Figure 3 .

[0032] Step 4, the welding surface 20 has one, and the point cloud data corresponding to the welding surface 20 is obtained.

[0033] The processing process of step 4 is as follows: Since the shape of the selected point cloud region 10 containing the welding surface is relatively regular, it can be approximated as including several planes. Use the Ransac plane segmentation algorithm to segment the outer surface of the selected point cloud region 10 containing the welding surface into several independent planes, remove useless point cloud data, as shown in Figure 4 , and retain the point cloud data corresponding to the welding surface 20, as shown in Figure 5 .

[0034] Step 5, preliminarily select path point cloud data in the point cloud data corresponding to the welding surface 20, and select several preliminary paths 40 from the welding surface 20.

[0035] Specifically, in step 5, the processing process of the welding surface 20 is as follows: Step 501, use principal component analysis algorithm to center the point cloud data corresponding to the welding surface 20.

[0036] ​Step 502, calculate the covariance matrix of the point cloud data corresponding to the welding surface 20, and obtain the eigenvectors corresponding to the maximum eigenvalue, the second maximum eigenvalue and the minimum eigenvalue of the covariance matrix. The eigenvector corresponding to the maximum eigenvalue is the long side direction of the welding surface 20, the eigenvector corresponding to the second maximum eigenvalue is the short side direction of the welding surface 20, and the eigenvector corresponding to the minimum eigenvalue is the normal direction of the welding surface 20. The welding surface 20 can be regarded as a rectangle, and the long side length and the short side length of the welding surface 20 are obtained by using the dot product formula. The height of the minimum bounding box of the point cloud corresponding to the welding surface 20 is obtained according to the minimum bounding box of the point cloud.

[0037] Step 503, a cutting surface 30 perpendicular to the welding surface 20 is generated on the welding surface 20. The intersection line between the cutting surface 30 and the welding surface 20 extends along the short side direction of the welding surface 20, and the distance between the intersection line and the short side of the welding surface 20 is a first preset distance, for example, the distance between them is half of the width of the welding bead or the width of the welding bead, which is set according to the requirement.

[0038] Specifically, step 503 includes generating four corner points of the cutting surface 30 by using the three directions and three sizes obtained in step 502, as shown in FIG. 3. Figure 6

[0039] The corner point generation method is as follows: first, the centroid point of the point cloud corresponding to the welding surface 20 is calculated, then the centroid point is translated by a distance of 1 / 2 of the long side length in the opposite direction of the long side direction of the welding surface 20, so that the centroid point reaches the midpoint of the short side at the left end of the welding surface 20, then the centroid point is translated by the first preset distance in the long side direction of the welding surface 20, then the centroid point is translated to the first point of the welding surface 20 by 1 / 2 of the short side length of the welding surface 20 multiplied by the positive short side direction, and then the centroid point is translated again by 1 / 2 of the height of the bounding box multiplied by the normal direction, so that the centroid point is located at the first corner point 301.

[0040] Similarly, the positions of the second corner point 302, the third corner point 303 and the fourth corner point 304 can be obtained. The position of the second corner point 302 is as follows: first, the centroid point of the point cloud corresponding to the welding surface 20 is calculated, then the centroid point is translated by a distance of 1 / 2 of the long side length in the opposite direction of the long side direction of the welding surface 20, so that the centroid point reaches the midpoint of the short side at the left end of the welding surface 20, then the centroid point is translated by the first preset distance in the long side direction of the welding surface 20, then the centroid point is translated to the first point of the welding surface 20 by 1 / 2 of the short side length of the welding surface 20 multiplied by the positive short side direction, and then the centroid point is translated again by 1 / 2 of the height of the bounding box multiplied by the normal direction, so that the centroid point is located at the first corner point 301.

[0041] ​The position of the third corner point 303: first, the centroid of the point cloud corresponding to the welding surface 20 is calculated, then the centroid is translated by a distance of 1 / 2 of the length of the long side in the opposite direction of the long side of the welding surface 20, so that the centroid reaches the midpoint of the short side at the left end of the welding surface 20, then the centroid is translated by the first preset distance in the long side direction of the welding surface 20; then the centroid is translated to the first point of the welding surface 20 by 1 / 2 of the length of the short side of the welding surface 20 multiplied by the short side direction in the positive direction; then the centroid is translated again to the position of the third corner point 303 by 1 / 2 of the height of the bounding box multiplied by the normal direction in the opposite direction.

[0042] The position of the fourth corner point 304: first, the centroid of the point cloud corresponding to the welding surface 20 is calculated, then the centroid is translated by a distance of 1 / 2 of the length of the long side in the opposite direction of the long side of the welding surface 20, so that the centroid reaches the midpoint of the short side at the left end of the welding surface 20, then the centroid is translated by the first preset distance in the long side direction of the welding surface 20; then the centroid is translated to the second point of the welding surface 20 by 1 / 2 of the length of the short side of the welding surface 20 multiplied by the short side direction in the opposite direction; then the centroid is translated again to the position of the fourth corner point 304 by 1 / 2 of the height of the bounding box multiplied by the normal direction in the opposite direction.

[0043] Step 504: using the dot product formula, the point cloud data within a second preset distance range from the current cutting surface 30 corresponding to the point cloud data of the welding surface 20 forms a preliminary path 40, and a group of path point cloud data is thus screened out. In this embodiment, the second preset distance can be 0.05 mm, and of course can be set to other values according to requirements.

[0044] Step 505: determining whether the remaining length of the welding surface 20 along the long side direction after being cut by the current cutting surface 30 is greater than the width of the weld, if yes, jumping to step 506, if not, ending. The width of the weld can be measured by process test.

[0045] Step 506: generating a next cutting surface 30 perpendicular to the welding surface 20 at a distance of a weld width from the current cutting surface 30 along the long side direction of the welding surface 20, and the adjacent two cutting surfaces 30 are parallel to each other; using the dot product formula, the point cloud data within a second preset distance range from the newly generated cutting surface 30 corresponding to the point cloud data of the welding surface 20 forms a preliminary path 40.

[0046] Specifically, step 506 comprises: translating the centroid of the to-be-welded surface 20 at the center of the current cutting surface 30 along the long edge direction of the to-be-welded surface 20 by a distance consistent with the width of the welding bead, and generating a next cutting surface 30 perpendicular to the to-be-welded surface 20 at the position after the translation. Specifically, the calculation process of the four corner points of the newly generated cutting surface 30 is similar to that of step 503, and will not be described here again.

[0047] Step 507: repeat steps 505 and 506 until the remaining length of the to-be-welded surface 20 along the long edge direction after cutting of the last cutting surface 30 is less than the width of the welding bead, thereby obtaining all the preliminary paths 40 corresponding to the to-be-welded surface 20, as shown in Figure 7

[0048] Step 6: processing the point cloud data corresponding to each preliminary path 40 obtained in step 5. For each preliminary path 40, a linear point cloud is obtained, and m points are selected from the linear point cloud as a group of path points, where m is a positive integer and m≥2.

[0049] Specifically, in step 6, the point cloud data corresponding to one of the preliminary paths 40 obtained in step 5 is processed to obtain a linear point cloud, and the processing process is as follows: the Ransac straight line segmentation algorithm is used to fit the point cloud data corresponding to the preliminary path 40 to obtain a fitting path 50 with higher precision, as shown in Figure 8 At this time, the fitting path 50 is represented in the form of a three-dimensional model, for example, an elongated cylinder. After the three-dimensional model corresponding to the fitting path 50 is point clouded, a corresponding linear point cloud is obtained. In this embodiment, five points are selected from the linear point cloud as a group of path points, which are used as path points for subsequent input to the welding robot. The five points are one at the head of the linear point cloud, one at 1 / 4 of the linear point cloud, one at 1 / 2 of the linear point cloud, one at 3 / 4 of the linear point cloud, and one at the tail of the linear point cloud.

[0050] Step 7: arranging the corresponding groups of path points on the to-be-welded surface 20 in the following manner: the serial numbers of the points in the path points with an odd number of groups are arranged in a forward direction; and the serial numbers of the points in the path points with an even number of groups are arranged in a reverse direction, as shown in Figure 9 Finally, the corresponding groups of path points on the to-be-welded surface 20 form a complete path, as shown in Figure 10

[0051] Step 8: substituting the TCP coordinates of the welding torch recorded in step 2 and the coordinates in the structured light camera coordinate system corresponding to the complete path obtained in step 7 into the coordinate conversion matrix obtained in step 1, converting the coordinates of the path points in the structured light camera coordinate system into coordinates in the world coordinate system, and obtaining the actual coordinate values of all the path points, thereby completing the path planning.

[0052] ​​The turbine blade wear-resistant layer welding path planning method based on point cloud processing involved in the present application can solve the problems of low welding precision, low intelligence and low automation of the engine turbine blade in the prior art. The above method has the following advantages: 1) The turbine blade wear-resistant layer welding process can realize automatic path planning, which can greatly improve the production efficiency; 2) The turbine blade welding process machining precision can be improved, and the error is within 0.05mm; 3) The placing requirements of the turbine blade in the production process are reduced, the intelligence is realized, and the production efficiency is improved.

[0053] The above is only the preferred specific embodiment of the present application, but the protection scope of the present application is not limited to this. Any skilled person in the art can make equivalent replacement or change according to the technical solution and concept of the present application within the technical range disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. A turbine blade wear-resistant layer welding path planning method based on point cloud processing, characterized by: The method comprises the following steps: Step 1, fixing a structured light camera on a welding torch, calibrating the structured light camera, obtaining a conversion matrix between a structured light camera coordinate system and a welding robot end coordinate system, and obtaining a conversion matrix between the welding robot end coordinate system and a world coordinate system; Step 2, scanning a turbine blade by using the structured light camera to obtain original three-dimensional point cloud data of the turbine blade, and recording coordinates of the welding torch in the welding robot end coordinate system; Step 3, preprocessing the original three-dimensional point cloud data obtained in step 2 to obtain a point cloud region containing point cloud data corresponding to a welding surface; Step 4, obtaining the point cloud data corresponding to the welding surface; Step 5, preliminarily screening path point cloud data in the point cloud data corresponding to the welding surface, and screening a plurality of preliminary paths from the welding surface; Step 6, processing point cloud data corresponding to each preliminary path obtained in step 5, and obtaining linear point cloud for each preliminary path, selecting m points in the linear point cloud as a group of path points, wherein m is a positive integer and m is greater than or equal to 2; Step 7, arranging a plurality of groups of path points corresponding to the welding surface in the following manner: arranging serial numbers of points in path points in an odd number of groups in a forward direction; arranging serial numbers of points in path points in an even number of groups in a reverse direction; and finally forming a complete path from the plurality of groups of path points corresponding to the welding surface; Step 8, substituting coordinates of the welding torch in the welding robot end coordinate system recorded in step 2 and coordinates of the complete path in the structured light camera coordinate system obtained in step 7 into the coordinate conversion matrix obtained in step 1, converting coordinates of the path points in the structured light camera coordinate system into coordinates in the world coordinate system, obtaining actual coordinate values of all path points, and thus completing path planning.

2. The method of claim 1, wherein: The preprocessing process involved in step 3 is as follows: Step 301, removing invalid points in the original point cloud data by using a straight-through filter to remove background point cloud data; Step 302, reconstructing the density of the point cloud obtained in step 301 by using a voxel filter; Step 303, segmenting the density-reconstructed point cloud into a plurality of point cloud regions by using a K-Means clustering method, and screening a point cloud region containing the welding surface.

3. The method of claim 2, wherein: The processing process of step 4 is as follows: using a Ransac plane segmentation algorithm to segment the outer surface of the screened point cloud region containing the welding surface into a plurality of independent surfaces, removing useless point cloud data, and retaining point cloud data corresponding to the welding surface.

4. The method of claim 3, wherein: In step 5, the processing process of the welding surface is as follows: Step 501, centralizing the point cloud data corresponding to the welding surface by using a principal component analysis algorithm; Step 502, calculate the covariance matrix of the point cloud data corresponding to the welding surface, and obtain the eigenvectors corresponding to the maximum eigenvalue, the second maximum eigenvalue and the minimum eigenvalue of the covariance matrix, wherein the eigenvector corresponding to the maximum eigenvalue is the long side direction of the welding surface, the eigenvector corresponding to the second maximum eigenvalue is the short side direction of the welding surface, and the eigenvector corresponding to the minimum eigenvalue is the normal direction of the welding surface; the welding surface is regarded as a rectangle, and the length of the long side and the length of the short side of the welding surface are obtained by using the dot product formula; and the height of the minimum bounding box corresponding to the point cloud of the welding surface is obtained according to the minimum bounding box. Step 503, generate a cutting surface perpendicular to the welding surface on the welding surface by using the three directions and three sizes obtained in step 502, wherein the intersection line of the cutting surface and the welding surface extends along the short side direction of the welding surface, and the distance between the intersection line and the short side of the left end of the welding surface is a first preset distance. Step 504, form a preliminary path by using the dot product formula and the point cloud data in the second preset distance range from the current cutting surface corresponding to the point cloud data of the welding surface. Step 505, determine whether the remaining length of the welding surface along the long side direction after being cut by the current cutting surface is greater than the width of the welding bead, if yes, jump to step 506, and if no, end. Step 506, generate a next cutting surface perpendicular to the welding surface along the long side direction of the welding surface at a distance of a welding bead width from the current cutting surface, and the adjacent two cutting surfaces are parallel to each other; form a preliminary path by using the dot product formula and the point cloud data in the second preset distance range from the newly generated cutting surface corresponding to the point cloud data of the welding surface. Step 507, repeat steps 505 and 506 until the remaining length of the welding surface along the long side direction after being cut by the last cutting surface is less than the width of the welding bead, and thus all preliminary paths corresponding to the welding surface are obtained.

5. The method of claim 4, wherein: In step 6, the point cloud data corresponding to one of the preliminary paths obtained in step 5 is processed to obtain a linear point cloud, and the processing process is as follows: a Ransac straight line segmentation algorithm is used to fit the point cloud data corresponding to the preliminary path to obtain a fitting path, at this time, the fitting path is in the form of a three-dimensional model, and the corresponding linear point cloud is obtained after the three-dimensional model corresponding to the fitting path is point clouded.

Images