A 3D Vision-Based Method for Recognizing and Analyzing the Chamfered Edge of a Support Plate

Abstract

This invention relates to the field of welded support plate reinforcement, specifically a support plate chamfer recognition and analysis method based on 3D vision. The specific steps are as follows: S1, reading point cloud; S2, point cloud preprocessing; S3, point cloud segmentation; S4, point cloud planar projection; S5, constructing a kd-tree; S6, edge detection; S7, line fitting and intersection calculation; S8, chamfer feature search; S9, whether it is a chamfer feature; S10, constructing a chamfer inner edge search direction group; S11, end. This invention is universal for support plates with chamfers. Even when the workpiece changes, as long as the support plate has a chamfer structure, it can accurately locate and identify it, and provide relevant parameters. This overcomes the problem that deep learning requires training on specific models and lacks universality.

Description

Technical Field

[0001] This invention relates to the field of welded support plate reinforcement, specifically a support plate chamfer recognition and analysis method based on 3D vision. Background Technology

[0002] In the field of welding, large or heavily stressed structures are typically reinforced with support plates. To ensure a more complete weld and continuity, the base of the support plate needs to be chamfered, usually in the form of a rounded arc or a straight line. In current robotic automated welding, the chamfered areas of the support plate are generally welded manually. This is because the weld seams in the chamfered areas have many bends and are relatively scattered, making it difficult to identify the chamfer parameters using visual technology and thus difficult to provide the required weld parameters for the chamfered areas. Summary of the Invention

[0003] To address the aforementioned issues, this invention proposes a support plate chamfer recognition and analysis method based on 3D vision.

[0004] The specific steps of the support plate chamfer recognition and analysis method based on 3D vision are as follows:

[0005] S1. Read the point cloud;

[0006] S2. Point cloud preprocessing: Select a preprocessing method based on the point cloud quality, then calculate the normal vectors of the point cloud and construct a kd-tree;

[0007] S3. Point cloud segmentation: Use point cloud segmentation methods to segment the plane;

[0008] S4. Point cloud planar projection: To improve accuracy, for the plane panel_i, select the points in the preprocessed point cloud that are closer to the plane and project them onto the plane.

[0009] S5. Construct a kd-tree;

[0010] S6. Edge Detection: Finding the edges of a planar point cloud: Preserving the edges of the planar point cloud using the BoundaryEstimation method;

[0011] S7. Line Fitting and Intersection Calculation: Perform RANSAC line fitting on the edge points and calculate the intersection points between the lines;

[0012] S8. Chamfer Feature Search: Construct a chamfer structure search vector, with the intersection of the lines as the center and the direction of the angle bisector of the lines as the vector, to construct four quadrant search directions;

[0013] S9. Is it a chamfer feature? When the point cloud distribution meets the initial conditions, that is, there are missing holes in the point cloud inside the chamfer plane, the angle bisectors of the intersecting straight lines and the two sides are calculated again to search for the point cloud distribution. If the three search results meet the chamfer rule, it is considered yes, and it is preliminarily judged that the plane meets the chamfer plane, and the chamfer parameters are calculated. If not, the current process ends and the next judgment process begins.

[0014] S10. Construct a group of chamfer inner edge search directions: Construct chamfer edge search directions in possible chamfer areas with a certain angle difference;

[0015] a) Chamfer Inner Edge Search: Search for the inner edge of the chamfer along the pre-defined search direction;

[0016] a1. Coordinate system transformation to calculate chamfer parameters: Using the angle as the x-axis and the distance from the inner edge point of the chamfer to the origin as the y-axis, transform the search results to a rectangular coordinate system, and then fit the transformation results to calculate the chamfer parameters.

[0017] b. Search the outer edge of the support plate; Use the chamfered edge search direction established earlier to search the outer edge of the support plate;

[0018] b1. Support plate thickness search: Search for the thickness according to the constructed support plate thickness search direction, that is, search along the normal vector direction of the search surface until there are no points, and record the search length;

[0019] b2. Coordinate system transformation to calculate thickness: Construct a search result coordinate system with the angle of the chamfered edge search direction as the x-axis and the search length in the thickness direction as the y-axis. Then fit and transform the coordinates to calculate the thickness.

[0020] S11, End.

[0021] In step S2, the preprocessing methods include removing NaN points, removing outliers, and downsampling point clouds.

[0022] In step S3, the point cloud segmentation method can be a point cloud segmentation method based on feature search and region growing, or for point clouds with relatively simple structures, RANSAC can be used directly to segment the plane.

[0023] In step S4, if the distance from a point to the plane is less than a threshold, projection is performed. The determination condition is as follows:

[0024] Abs(A*x+B*y+C*z+D) <dis

[0025] Where dis is the distance threshold;

[0026] Point projection, the coordinates of the projected point are:

[0027] point_out-(point_out-point_plane).dot(plane_normal)*plane_normal

[0028] Where (point_out-point_plane).dot(plane_normal) is the projection distance of the vector formed by the two points point_out and point_plane onto the vector plane_normal, which is the distance from point_out to the plane.

[0029] In step S8, the point cloud distribution is determined by searching along the constructed search direction, and the iterative calculation formula for the search point position is as follows:

[0030] pij=pi+vp*j*step

[0031] Where j is the number of iterations and step is the search step size;

[0032] When searching for the inner edge of the bevel, record the nearest distance to the point cloud found;

[0033] When searching the outer edge, record the farthest position where the number of search point clouds is not empty.

[0034] In step S9, after the chamfer condition is met in the plane, it is also necessary to verify that the chamfer meets the structural requirements. There are two supporting surfaces on both sides of the chamfer, so the number of point clouds in the vertical direction of the chamfer plane is more than that in the direction of the chamfer plane. After the chamfer condition is met, the chamfer parameter fitting process is performed.

[0035] In step S10, the method for constructing the search direction is as follows:

[0036] v3=cos(Δa*j)*v1+sin(Δa*j)*v2

[0037] Where v1 and v2 are two intersecting edge lines, Δa is the difference in the interval angle, and j is the number of iterations.

[0038] In step S10a1, the result after coordinate system transformation is fitted with a straight line using the RANSAC straight line fitting method or with a parabola using the least squares method:

[0039] a) For a circular arc chamfer, after coordinate system transformation, the search result is fitted to a straight line, and the distance of the straight line from the x-axis is the radius of the circular arc chamfer;

[0040] b) For a straight chamfer, the fitting result after coordinate system transformation is a concave parabola. The intersection of the original coordinate fitting line corresponding to the two endpoints of the parabola and the two axes of the chamfer is the position coordinate of the two sides of the straight chamfer.

[0041] In step S10b, after the edge is found, the support plate thickness search direction is constructed at the edge. That is, at each edge point, the search is performed along the normal direction of the support surface or the opposite direction of the normal, thus constructing the support plate thickness search direction.

[0042] In step S10, b2, the RANSAC linear fitting method is used to fit the coordinate system transformation result, and the height of the straight line from the x-axis is the thickness of the support plate.

[0043] The beneficial effects of this invention are: this invention is universal for support plates with chamfers. When the workpiece changes, as long as there is a chamfered structure on the support plate, it can accurately locate and identify it and provide relevant parameters, thus overcoming the problem that deep learning requires training for a specific model and lacks universality. Attached Figure Description

[0044] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0045] Figure 1 This is a schematic diagram of the process structure of the present invention;

[0046] Figure 2 This is a schematic diagram of a point projected onto a plane according to the present invention;

[0047] Figure 3 This is a schematic diagram of the planar extraction method of the present invention;

[0048] Figure 4 This is a schematic diagram of edge detection according to the present invention;

[0049] Figure 5 This is a schematic diagram of linear fitting according to the present invention;

[0050] Figure 6 This is a schematic diagram illustrating the search direction for constructing the present invention;

[0051] Figure 7 This is a schematic diagram of the angle bisector search of the present invention;

[0052] Figure 8 This is a schematic diagram of the secondary search in this invention;

[0053] Figure 9 This is a schematic diagram of the chamfered structure of the present invention;

[0054] Figure 10 This is a schematic diagram of the chamfered structure verification model of the present invention;

[0055] Figure 11 This is a schematic diagram illustrating the search direction for constructing the chamfered region according to the present invention;

[0056] Figure 12This is a schematic diagram of the chamfered edge search of the present invention;

[0057] Figure 13 This is a schematic diagram of the coordinate system transformation of the present invention;

[0058] Figure 14 This is a schematic diagram of linear fitting according to the present invention;

[0059] Figure 15 This is a schematic diagram of the outer edge lines of the search support plate of the present invention;

[0060] Figure 16 This is a schematic diagram of the outer edge of the search support plate of the present invention;

[0061] Figure 17 This is a schematic diagram of the search thickness according to the present invention;

[0062] Figure 18 This is a schematic diagram illustrating the chamfer recognition effect of the present invention. Detailed Implementation

[0063] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below.

[0064] like Figures 1 to 18 As shown, the specific steps of the support plate chamfer recognition and analysis method based on 3D vision are as follows:

[0065] S1. Read the point cloud;

[0066] S2. Point cloud preprocessing: Select a preprocessing method based on the point cloud quality, then calculate the normal vectors of the point cloud and construct a kd-tree;

[0067] S3. Point cloud segmentation: Use point cloud segmentation methods to segment the plane;

[0068] S4. Point cloud planar projection: To improve accuracy, for the plane panel_i, select the points in the preprocessed point cloud that are closer to the plane and project them onto the plane.

[0069] S5. Construct a kd-tree;

[0070] S6. Edge Detection: Finding the edges of a planar point cloud: Using the BoundaryEstimation method to preserve the edges of the planar point cloud, the result is as follows. Figure 4 As shown;

[0071] S7. Line Fitting and Intersection Calculation: Perform RANSAC line fitting on the edge points and calculate the intersection points between the lines;

[0072] S8. Chamfer Feature Search: Construct a chamfer structure search vector, such as... Figure 6As shown, with the intersection of the lines as the center and the direction of the angle bisector of the line as the vector, four quadrant search directions are constructed;

[0073] S9. Is it a chamfer feature? (e.g.) Figure 7 As shown, when the point cloud distribution meets the initial condition, i.e., there are missing holes in the point cloud inside the chamfered plane, the angle bisectors of the intersecting straight lines and both sides are calculated again to search for the point cloud distribution; as shown... Figure 8 As shown, if the three search results match the chamfering rule, it is considered "yes," and the plane is initially judged to be a chamfering plane, and the chamfering parameter calculation continues; otherwise, the current process ends and the next judgment process begins.

[0074] S10. Construct a group of chamfer inner edge search directions: Construct chamfer edge search directions in possible chamfer areas with a certain angle difference;

[0075] b. Bevel inner edge search: Search for the inner edge of the bevel along the pre-defined search direction, such as... Figure 12 As shown;

[0076] a1. Coordinate system transformation to calculate chamfer parameters: Using the angle as the x-axis and the distance from the inner edge of the chamfer to the origin as the y-axis, transform the search results to a rectangular coordinate system. Then, fit the transformed results to calculate the chamfer parameters. Figure 13 As shown;

[0077] b. Search the outer edge of the support plate; Use the chamfered edge search direction established earlier to search the outer edge of the support plate;

[0078] b1. Support plate thickness search: Search for the thickness according to the constructed support plate thickness search direction, that is, search along the search normal vector until there are no points, and record the search length;

[0079] b2. Coordinate system transformation to calculate thickness: Construct a search result coordinate system with the angle of the chamfered edge search direction as the x-axis and the search length in the thickness direction as the y-axis. Then fit and transform the coordinates to calculate the thickness.

[0080] S11, End.

[0081] In step S2, the preprocessing methods include removing NaN points, removing outliers, and downsampling point clouds.

[0082] This invention is universal for support plates with chamfers. When the workpiece changes, it can accurately locate and identify the support plate as long as it has a chamfer structure and provide relevant parameters. This overcomes the problem that deep learning requires training on a specific model and lacks universality.

[0083] In step S3, the point cloud segmentation method can be a point cloud segmentation method based on feature search and region growing, or for point clouds with relatively simple structures, RANSAC can be used directly to segment the plane.

[0084] In step S4, if the distance from a point to the plane is less than a threshold, projection is performed. The determination condition is as follows:

[0085] Abs(A*x+B*y+C*z+D) <dis

[0086] Where dis is the distance threshold;

[0087] Point projection onto a plane, such as Figure 2 As shown, the coordinates of the projected point are:

[0088] point_out-(point_out-point_plane).dot(plane_normal)*plane_normal

[0089] Where (point_out-point_plane).dot(plane_normal) is the projection distance of the vector formed by the two points point_out and point_plane onto the vector plane_normal, which is the distance from point_out to the plane.

[0090] In step S8, the point cloud distribution is determined by searching along the constructed search direction, and the iterative calculation formula for the search point position is as follows:

[0091] pij=pi+vp*j*step

[0092] Where j is the number of iterations and step is the search step size;

[0093] When searching for the inner edge of the bevel, record the nearest distance to the point cloud found;

[0094] When searching the outer edge, record the farthest position where the number of search point clouds is not empty.

[0095] In step S9, after the chamfer condition is met in the plane, it is also necessary to verify that the chamfer meets the structural requirements, such as... Figure 10 As shown, there are two supporting surfaces on both sides of the chamfer. Therefore, the number of point clouds in the direction perpendicular to the chamfer plane is greater than that in the direction of the chamfer plane. After the chamfer conditions are met, the chamfer parameters are fitted.

[0096] In step S10, as described above... Figure 11 As shown, the method for constructing the search direction is as follows:

[0097] v3=cos(Δa*j)*v1+sin(Δa*j)*v2

[0098] Where v1 and v2 are two intersecting edge lines, Δa is the difference in the interval angle, and j is the number of iterations.

[0099] In step S10a1, the result after coordinate system transformation is fitted with a straight line using the RANSAC straight line fitting method or with a parabola using the least squares method, such as... Figure 14 As shown:

[0100] a) For a circular arc chamfer, after coordinate system transformation, the search result is fitted to a straight line, and the distance of the straight line from the x-axis is the radius of the circular arc chamfer;

[0101] b) For a straight chamfer, the fitting result after coordinate system transformation is a concave parabola. The intersection of the original coordinate fitting line corresponding to the two endpoints of the parabola and the two axes of the chamfer is the position coordinate of the two sides of the straight chamfer.

[0102] In step S10b, as described above... Figure 15 As shown, after finding the edge, a search direction for the support plate thickness is constructed at the edge. That is, at each edge point, the search is performed along the normal direction of the support surface or in the opposite direction. The constructed search direction for the support plate thickness is as follows: Figure 15 As shown.

[0103] In step S10, b2, the fitting result is as follows: Figure 17 As shown, the RANSAC linear fitting method is used to fit the coordinate system transformation results, and the height of the line from the x-axis is the thickness of the support plate.

[0104] This invention implements methods for chamfer region search, chamfer parameter calculation, support structure verification, and support plate thickness calculation, which can accurately provide chamfer position and parameters, and provide weld parameters for the chamfer region to be welded.

[0105] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely prisms of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A method for identifying and analyzing the chamfer of a support plate based on 3D vision, characterized in that: The specific steps are as follows: S1. Read the point cloud; S2. Point cloud preprocessing: Select a preprocessing method based on the point cloud quality, then calculate the normal vectors of the point cloud and construct a kd-tree; S3. Point cloud segmentation: Use point cloud segmentation methods to segment the plane; S4. Point cloud planar projection: To improve accuracy, for the plane panel_i, select the points in the preprocessed point cloud that are closer to the plane and project them onto the plane. S5. Construct a kd-tree; S6. Edge Detection: Finding the edges of a planar point cloud: Preserving the edges of the planar point cloud using the BoundaryEstimation method; S7. Line Fitting and Intersection Calculation: Perform RANSAC line fitting on the edge points and calculate the intersection points between the lines; S8. Chamfer Feature Search: Construct a chamfer structure search vector, with the intersection of the lines as the center and the direction of the angle bisector of the lines as the vector, to construct four quadrant search directions; S9. Is it a chamfer feature? When the point cloud distribution meets the initial conditions, that is, there are missing holes in the point cloud inside the chamfer plane, the angle bisectors of the intersecting straight lines and the two sides are calculated again to search for the point cloud distribution. If the three search results meet the chamfer rule, it is considered yes, and it is preliminarily judged that the plane meets the chamfer plane, and the chamfer parameters are calculated. If not, the current process ends and the next judgment process begins. S10. Construct a group of chamfer inner edge search directions: Construct chamfer edge search directions in possible chamfer areas with a certain angle difference; Chamfer inner edge search: Search for the inner edge of the chamfer along the pre-defined search direction; a1. Coordinate system transformation to calculate chamfer parameters: Using the angle as the x-axis and the distance from the inner edge point of the chamfer to the origin as the y-axis, transform the search results to a rectangular coordinate system, and then fit the transformation results to calculate the chamfer parameters. b. Search the outer edge of the support plate; Use the chamfered edge search direction established earlier to search the outer edge of the support plate; b1. Support plate thickness search: Search for the thickness according to the constructed support plate thickness search direction, that is, search along the normal vector direction of the search surface until there are no points, and record the search length; b2. Coordinate system transformation to calculate thickness: Construct a search result coordinate system with the angle of the chamfered edge search direction as the x-axis and the search length in the thickness direction as the y-axis. Then fit and transform the coordinates to calculate the thickness. S11, End.

2. The support plate chamfer recognition and analysis method based on 3D vision according to claim 1, characterized in that: In step S2, the preprocessing methods include removing NaN points, removing outliers, and downsampling point clouds.

3. The support plate chamfer recognition and analysis method based on 3D vision according to claim 1, characterized in that: In step S3, the point cloud segmentation method uses a point cloud segmentation method based on feature search and region growing, or for point clouds with relatively simple structures, RANSAC is used directly to segment the plane.

4. The support plate chamfer recognition and analysis method based on 3D vision according to claim 1, characterized in that: In step S4, if the distance from a point to the plane is less than a threshold, projection is performed. The determination condition is as follows: Abs(A * x + B * y + C * z + D) < dis; Where dis is the distance threshold; Point projection, the coordinates of the projected point are: point_out - (point_out- point_plane).dot(plane_normal) * plane_normal; Where (point_out - point_plane).dot(plane_normal) is the projection distance of the vector formed by the two points point_out and point_plane onto the vector plane_normal, which is the distance from point_out to the plane; point_out refers to a point outside the plane; point_plane refers to a point on the plane; and plane_normal refers to the plane normal vector.

5. The support plate chamfer recognition and analysis method based on 3D vision according to claim 1, characterized in that: In step S9, after the chamfer condition is met in the plane, it is also necessary to verify that the chamfer meets the structural requirements. There are two supporting surfaces on both sides of the chamfer, so the number of point clouds in the vertical direction of the chamfer plane is more than that in the direction of the chamfer plane. After meeting the first two conditions, the chamfer plane search feature condition and the chamfer structure feature condition are satisfied, and then the chamfer parameter fitting process is performed.

6. The support plate chamfer recognition and analysis method based on 3D vision according to claim 1, characterized in that: In step S10, the method for constructing the search direction is as follows: v3 = cos(Δa*j) * v1 + sin(Δa*j) * v2; Where v1 and v2 are two intersecting edge lines, Δa is the difference in interval angle, and j is the number of iterations.

7. The support plate chamfer recognition and analysis method based on 3D vision according to claim 1, characterized in that: In step S10a1, the result after coordinate system transformation is fitted with a straight line using the RANSAC straight line fitting method or with a parabola using the least squares method: a) For a circular chamfer, after coordinate system transformation, the search result is fitted to a straight line, and the distance of the straight line from the x-axis is the radius of the circular chamfer; b) For a straight chamfer, the fitting result after coordinate system transformation is a concave parabola. The intersection of the original coordinate fitting line corresponding to the two endpoints of the parabola and the two axes of the chamfer is the position coordinate of the two sides of the straight chamfer.

8. The support plate chamfer recognition and analysis method based on 3D vision according to claim 1, characterized in that: In step S10b, after the edge is found, the support plate thickness search direction is constructed at the edge. That is, at each edge point, the search is performed along the normal direction of the support surface or the opposite direction of the normal, thus constructing the support plate thickness search direction.

9. The support plate chamfer recognition and analysis method based on 3D vision according to claim 1, characterized in that: In step S10, b2, the RANSAC linear fitting method is used to fit the coordinate system transformation result, and the height of the straight line from the x-axis is the thickness of the support plate.

Images