A double-robot arm steel beam automatic detection system and method

Abstract

The application discloses a kind of double mechanical arm steel beam automatic detection system and method, the detection system includes linear rail device;Movable gantry component, including transverse connecting beam and being arranged in the sliding connection of guide rail mobile slide, vertical extension support column is installed on mobile slide, the both ends of transverse connecting beam are connected with support column;Workpiece support positioning device;Control component;Wherein, the lower of transverse connecting beam is also provided with the first inverted six-axis mechanical arm and the second inverted six-axis mechanical arm that are arranged in space staggered, the end of first inverted six-axis mechanical arm and second inverted six-axis mechanical arm are equipped with structured light scanning device, control component is electrically connected with linear rail device, first inverted six-axis mechanical arm, second inverted six-axis mechanical arm and structured light scanning device respectively.The application can solve the problems of low detection efficiency and insufficient automation in the prior art.

Description

Technical Field

[0001] This invention belongs to the field of steel structure inspection technology, and particularly relates to an automatic inspection system and method for steel beams with dual robotic arms. Background Technology

[0002] Welded H-beams and welded solid web steel beams are core load-bearing components of modern steel structure buildings, bridges and industrial plants. Their dimensional accuracy and weld quality are directly related to structural safety, assembly effect and service life, and are key links in quality control during production and manufacturing. Currently, on-site factory inspections are still mainly conducted manually, relying on tools such as steel rulers, squares, and feeler gauges to measure manually, combined with visual inspection to determine weld defects. This method generally suffers from problems such as low efficiency, strong subjectivity, incomplete inspection, and lack of data traceability. For ultra-long steel beams up to 12m in length, manual inspection can only be carried out by sampling key parameters. It is difficult to continuously measure the form and position deviations such as bending, twisting, web flatness, and flange perpendicularity. The measurement error is large and cannot meet the high-precision acceptance requirements of national standards. Weld inspection relies on manual experience, and it is easy to miss or misjudge minor defects such as porosity, undercut, and abnormal reinforcement height. It is also impossible to accurately and quantitatively detect weld leg size, weld width, etc. In recent years, structured light 3D scanning technology has begun to be applied to component inspection. However, most existing solutions are portable handheld devices with low automation. They require manual adjustment of the station position, making it difficult to achieve stable full-coverage scanning of ultra-long components. Point cloud stitching errors are large. At the same time, most systems only support single functions such as size measurement or weld inspection, and cannot complete all-item inspection in an integrated manner. Furthermore, they lack stability in harsh environments such as workshop dust and welding spatter, making it difficult to adapt to the needs of industrial mass production. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of the prior art by providing a dual-robotic arm automatic steel beam detection system and method, which solves the problems of low detection efficiency and insufficient automation in the prior art.

[0004] The objective of this invention is achieved through the following technical solution: A dual-robotic arm automatic steel beam inspection system includes: A linear ground rail device includes two guide rails arranged opposite each other, both of which extend along the length of a steel beam. Each guide rail is also provided with a rack and a drive gear assembly that meshes with the rack. A mobile gantry assembly includes a transverse connecting beam and a movable slide block that is slidably connected to the guide rail. A vertically extending support column is mounted on the movable slide block, and both ends of the transverse connecting beam are connected to the support column. The workpiece support and positioning device includes multiple support brackets arranged at intervals along the length of the steel beam, and each support bracket is provided with a mechanical stop. Control components; Below the transverse connecting beam, there are a first inverted six-axis robotic arm and a second inverted six-axis robotic arm arranged in a spatially staggered manner. The ends of the first and second inverted six-axis robotic arms are equipped with structured light scanning devices. The control components are electrically connected to the linear ground rail device, the first inverted six-axis robotic arm, the second inverted six-axis robotic arm, and the structured light scanning device, respectively.

[0005] In one embodiment, the linear ground rail device further includes a plurality of mechanical positioning pins spaced apart along the length of the steel beam, and the movable gantry assembly has positioning elements corresponding to the mechanical positioning pins to perform segmented precise positioning of the movable gantry assembly.

[0006] In one embodiment, the transverse connecting beam is a hollow beam structure with a cable routing channel inside.

[0007] In one embodiment, the structured light scanning device includes a structured light projector and an industrial camera, and the structured light scanning device is fixed to the end effector of a robotic arm via an angle-adjustable mounting bracket.

[0008] The present invention also provides an automatic detection method for steel beams with dual robotic arms, comprising the following steps: Step S1: Place the steel beam on the workpiece support and positioning device and complete the initial positioning; Step S2: Use the linear ground rail device to drive the moving gantry assembly to the target position and position it using mechanical positioning pins; Step S3: Control the first inverted six-axis robotic arm and the second inverted six-axis robotic arm to move to the preset scanning pose; Step S4: Obtain point cloud data of the steel beam surface from multiple perspectives using a structured light scanning device; Step S5: Denoise and downsample the point cloud data; Step S6: Perform collaborative calibration and point cloud stitching between the first inverted six-axis robotic arm and the second inverted six-axis robotic arm to obtain a complete steel beam surface model; Step S7: Output the steel beam dimensions and weld quality inspection results.

[0009] In one embodiment, in step S4, a structured light stripe is projected onto the surface of the steel beam using a structured light scanning device, and an industrial camera is used to acquire stripe deformation images to calculate the three-dimensional point cloud coordinates of the steel beam surface.

[0010] in, The depth value of the target point. For industrial camera focal length, This represents the baseline distance between the structured light emitter and the industrial camera. This represents the disparity value of the striped image.

[0011] In one implementation, step S5 involves denoising the point cloud data, including: Calculate the average distance between each point in the input point cloud and its k nearest neighbors:

[0012] in, For the first The average neighborhood distance of each point For the first The point and the first Euclidean distance between neighboring points; Calculate the Gaussian distribution of the average distance between all points, and then calculate the mean μ and standard deviation σ. Set distance threshold: ; Where α is the threshold coefficient; The system iterates through all points in the input point cloud. When the average neighborhood distance of a point is greater than a distance threshold, the point is identified as an outlier and deleted to achieve noise reduction of the point cloud data.

[0013] In one implementation, step S5 involves downsampling the point cloud data, including: Based on the three-dimensional coordinate set of the point cloud data, calculate the maximum and minimum values ​​in the x, y, and z directions, and establish the minimum bounding box of the point cloud; Set the voxel mesh edge length r, and divide the three-dimensional space into voxels; Calculate the index value of each point in the voxel grid:

[0014] in, , , These are the voxel indices of the point in the x, y, and z directions, respectively; Points within the same voxel mesh are grouped into a single set, and the centroid coordinates of all points within that voxel are calculated:

[0015] in, The coordinates of the voxel center point; The original point cloud data within the voxel is replaced with the coordinates of the voxel center point for downsampling processing.

[0016] In one embodiment, step S6 further includes: Step S601: Establish the spatial transformation relationship between the coordinate systems of the two robotic arms' bases: Control the first inverted six-axis robotic arm and the second inverted six-axis robotic arm to move to the preset calibration postures, so that the structured light scanning devices installed at the ends of the two robotic arms simultaneously observe a fixed common calibration target in space. The common calibration target is a planar checkerboard target or a three-dimensional spherical target. Record the pose matrices of the ends of the two robotic arms in their respective base coordinate systems. and The target pose matrix relative to the coordinate systems of the two robotic arms' ends is extracted using the target image acquired by the structured light scanning device. and Solve for the homogeneous transformation matrix between the coordinate systems of the two robotic arms' bases. ,satisfy: ; Step S602: Acquire point cloud data scanned by the two robotic arms: The first inverted six-axis robotic arm and the second inverted six-axis robotic arm scan different areas of the steel beam according to a preset trajectory, and acquire two pieces of original point cloud data through their respective structured light scanning devices. and And record the end-effector pose at each scanning moment; Step S603: Perform coarse point cloud stitching based on calibration results: using the base coordinate system transformation matrix obtained in step S601. The point cloud acquired by the first inverted six-axis robotic arm The point cloud is obtained by transforming the tool coordinate system to the base coordinate system of the second inverted six-axis robotic arm. The transformation relationship is as follows: ; This makes two point clouds and Under the same global coordinate system, preliminary spatial alignment is completed; Step S604: Optimize the point cloud stitching result using a sampling consistency initial registration algorithm based on fast point feature histograms: extract point clouds respectively. and The FPFH feature descriptor for each point in the source point cloud Several sampling points are randomly selected from the target point cloud based on feature similarity. Find the corresponding point in the matrix, construct candidate transformation matrices, and evaluate the optimal initial transformation matrix based on the distance error. To reduce the positional deviation between the two point clouds; Step S605: Use the generalized iterative nearest neighbor algorithm to perform precise registration of the point cloud.

[0017] In one implementation, step S605 involves using a generalized iterative nearest neighbor algorithm for precise point cloud registration, including: Given a source point cloud P and a target point cloud Q, initialize the rigid transformation matrix T; Calculate the covariance matrix in the neighborhood of each point to describe the local geometric features of the point cloud; The correspondence between the source point cloud and the target point cloud is established using the nearest neighbor search algorithm; Construct the error function:

[0018] in, and These are the covariance matrices of the source point cloud and the target point cloud, respectively. Minimize the error function iteratively Continuously optimize the rigid transformation matrix This continues until the error converges, achieving accurate point cloud registration.

[0019] The beneficial effects of this invention are as follows: By working in concert with the dual structured light scanning devices on the first and second inverted six-axis robotic arms, the scanning speed is significantly improved. At the same time, the entire inspection process runs automatically without manual intervention. It is applicable to the inspection of steel beams of different sizes, and can complete the fully automatic three-dimensional scanning and dimensional inspection of precast steel beams. The computer is used to perform point cloud denoising, point cloud registration and automatic dimensional calculation. It can automatically complete the qualification judgment according to the steel structure specifications and finally output a complete inspection report. Attached Figure Description

[0020] The invention will now be described in more detail with reference to embodiments and the accompanying drawings. Figure 1 A schematic diagram of the system structure of the present invention is shown; Figure 2 A flowchart illustrating the detection method of the present invention is shown; In the accompanying drawings, the same parts use the same reference numerals. The drawings are not to scale.

[0021] Figure label: 1-Linear ground rail device, 2-Moving gantry assembly, 3-First inverted six-axis robotic arm, 4-Second inverted six-axis robotic arm, 5-Structured light scanning device, 6-Workpiece support and positioning device, 7-Control components. Detailed Implementation

[0022] The invention will now be further described with reference to the accompanying drawings.

[0023] This invention provides an automatic inspection system for steel beams with dual robotic arms, such as... Figure 1 As shown, it includes: The linear ground rail device 1 includes two guide rails arranged opposite each other, both of which extend along the length of the steel beam. The guide rails are also provided with racks and drive gear assemblies that mesh with the racks. The movable gantry assembly 2 includes a transverse connecting beam and a movable slide block that is slidably connected to the guide rail. A vertically extending support column is installed on the movable slide block, and both ends of the transverse connecting beam are connected to the support column. The workpiece support and positioning device 6 includes multiple support brackets arranged at intervals along the length of the steel beam, and each support bracket is equipped with a mechanical stop. Control component 7; Below the transverse connecting beam, there are a first inverted six-axis robotic arm 3 and a second inverted six-axis robotic arm 4 arranged in a spatially staggered manner. The ends of the first inverted six-axis robotic arm 3 and the second inverted six-axis robotic arm 4 are equipped with structured light scanning devices 5. The control component 7 is electrically connected to the linear ground rail device 1, the first inverted six-axis robotic arm 3, the second inverted six-axis robotic arm 4 and the structured light scanning device 5, respectively. It should be noted that in this embodiment, the scanning speed is significantly improved by the coordinated operation of the dual structured light scanning device 5 on the first inverted six-axis robotic arm 3 and the second inverted six-axis robotic arm 4. At the same time, the entire detection process runs automatically without manual intervention and can be applied to the detection of steel beams of different sizes.

[0024] Specifically, such as Figure 1 As shown, the linear ground rail device 1 also includes a plurality of mechanical positioning pins spaced apart along the length of the steel beam, and the moving gantry assembly 2 has positioning parts corresponding to the mechanical positioning pins to perform segmented precise positioning of the moving gantry assembly 2. In this embodiment, the transverse connecting beam is a hollow beam structure with a cable routing channel inside for arranging robotic arm cables, sensor lines and signal lines. In this embodiment, the vertical support column of the mobile gantry assembly 2 adopts a box-type steel structure or a profile welded structure, and is equipped with internal reinforcing ribs to improve the overall rigidity and vibration resistance. Furthermore, both the first inverted six-axis robotic arm 3 and the second inverted six-axis robotic arm 4 adopt an upper base mounting form, and their mounting flanges are fixed to the lower surface of the robotic arm mounting base by high-strength bolts, with the robotic arm body extending downward to form an inverted working structure. In one embodiment, the structured light scanning device 5 includes a structured light projector and an industrial camera, and is fixed to the end effector of the robotic arm by an angle-adjustable mounting bracket for adjusting the scanning posture. In one embodiment, the workpiece support and positioning device 6 includes a plurality of support brackets arranged at intervals along the length of the steel beam, each support bracket being provided with a V-shaped or flat support surface, and a mechanical stop for restricting the axial and lateral movement of the steel beam.

[0025] In one embodiment, the bottom of the support bracket is provided with an adjustable height mechanism, which is used to adapt to the support height of steel beams of different specifications.

[0026] In one embodiment, the control component 7 includes a PLC control unit, a robotic arm controller, and an industrial computer. The PLC control unit is used to execute motion control logic, the robotic arm controller is used to control the first inverted six-axis robotic arm 3 and the second inverted six-axis robotic arm 4, and the industrial computer is used to complete point cloud data processing and detection analysis.

[0027] The present invention also provides an automatic detection method for steel beams with dual robotic arms, comprising the following steps: Step S1: Place the steel beam on the workpiece support and positioning device and complete the initial positioning; Step S2: Use the linear ground rail device to drive the moving gantry assembly to the target position and position it using mechanical positioning pins; Step S3: Control the first inverted six-axis robotic arm and the second inverted six-axis robotic arm to move to the preset scanning pose; Step S4: Obtain point cloud data of the steel beam surface from multiple perspectives using a structured light scanning device; Specifically, a structured light beam is projected onto the surface of the steel beam using a structured light scanning device, and an industrial camera is used to capture images of the stripe deformation to calculate the three-dimensional point cloud coordinates of the steel beam surface.

[0028] in, The depth value of the target point. For industrial camera focal length, This represents the baseline distance between the structured light emitter and the industrial camera. The parallax value of the stripe image; that is, by fusing point cloud data acquired under multiple scanning poses, multi-view point cloud data of the steel beam surface is formed. Step S5: Denoise and downsample the point cloud data; Specifically, the point cloud data is denoised, including: Calculate the average distance between each point in the input point cloud and its k nearest neighbors:

[0029] in, For the first The average neighborhood distance of each point For the first The point and the first Euclidean distance between neighboring points; Calculate the Gaussian distribution of the average distance between all points, and then calculate the mean μ and standard deviation σ. Set distance threshold: ; Where α is the threshold coefficient; Traverse each point in the input point cloud. When the average neighborhood distance of a point is greater than the distance threshold, the point is identified as an outlier noise point and deleted to achieve point cloud data noise reduction. Downsampling of point cloud data includes: Based on the three-dimensional coordinate set of the point cloud data, calculate the maximum and minimum values ​​in the x, y, and z directions, and establish the minimum bounding box of the point cloud; Set the voxel mesh edge length r, and divide the three-dimensional space into voxels; Calculate the index value of each point in the voxel grid:

[0030] in, , , These are the voxel indices of the point in the x, y, and z directions, respectively; Points within the same voxel mesh are grouped into a single set, and the centroid coordinates of all points within that voxel are calculated:

[0031] in, The coordinates of the voxel center point; The original point cloud data within the voxel is replaced with the coordinates of the voxel center point for downsampling processing; Step S6: Perform collaborative calibration and point cloud stitching between the first inverted six-axis robotic arm and the second inverted six-axis robotic arm to obtain a complete steel beam surface model; Specifically, step S6 includes: Step S601: Establish the spatial transformation relationship between the coordinate systems of the two robotic arms' bases: Control the first inverted six-axis robotic arm and the second inverted six-axis robotic arm to move to the preset calibration postures, so that the structured light scanning devices installed at the ends of the two robotic arms simultaneously observe a fixed common calibration target in space. The common calibration target is a planar checkerboard target or a three-dimensional spherical target. Record the pose matrices of the ends of the two robotic arms in their respective base coordinate systems. and The target pose matrix relative to the coordinate systems of the two robotic arms' ends is extracted using the target image acquired by the structured light scanning device. and Solve for the homogeneous transformation matrix between the coordinate systems of the two robotic arms' bases. ,satisfy: ; Among them, the common calibration target is a three-dimensional calibration block with a high-precision ceramic ball array attached to its surface. The distance between the center of the balls is calibrated by a coordinate measuring machine, and the error of the center of the balls is less than 0.01mm. The structured light scanning device extracts the center of the balls by spherical fitting method. Combined with the pose of the end of the robotic arm, the transformation matrix of the coordinate system of the dual robotic arm base is solved by the least squares method. The random error is reduced by averaging multiple measurements. Step S602: Acquire point cloud data scanned by the two robotic arms: The first inverted six-axis robotic arm and the second inverted six-axis robotic arm scan different areas of the steel beam according to a preset trajectory, and acquire two pieces of original point cloud data through their respective structured light scanning devices. and And record the end-effector pose at each scanning moment; Step S603: Perform coarse point cloud stitching based on calibration results: using the base coordinate system transformation matrix obtained in step S601. The point cloud acquired by the first inverted six-axis robotic arm The point cloud is obtained by transforming the tool coordinate system to the base coordinate system of the second inverted six-axis robotic arm. The transformation relationship is as follows: ; This makes two point clouds and Under the same global coordinate system, preliminary spatial alignment is completed; Step S604: Optimize the point cloud stitching result using a sampling consistency initial registration algorithm based on fast point feature histograms: extract point clouds respectively. and The FPFH feature descriptor for each point in the source point cloud Several sampling points are randomly selected from the target point cloud based on feature similarity. Find the corresponding point in the matrix, construct candidate transformation matrices, and evaluate the optimal initial transformation matrix based on the distance error. To reduce the positional deviation between the two point clouds; In step S604, the calculation radius of the fast point feature histogram descriptor is... The maximum number of iterations in the sampling consistency initial registration algorithm is set to 10,000, and the distance error threshold is set to 1.5 times the average spacing of the point cloud, to ensure the robustness and efficiency of coarse stitching. Step S605: Perform precise registration of the point cloud using the generalized iterative nearest neighbor algorithm; In step S605, the generalized iterative nearest neighbor algorithm is used for accurate point cloud registration, including: Given a source point cloud P and a target point cloud Q, initialize the rigid transformation matrix T; Calculate the covariance matrix in the neighborhood of each point to describe the local geometric features of the point cloud; The correspondence between the source point cloud and the target point cloud is established using the nearest neighbor search algorithm; Construct the error function:

[0032] in, and These are the covariance matrices of the source point cloud and the target point cloud, respectively. Minimize the error function iteratively Continuously optimize the rigid transformation matrix Continue until the error converges, achieving accurate point cloud registration; Step S7: Output the steel beam dimensions and weld quality inspection results.

[0033] It should be noted that in this embodiment, the scanning speed is significantly improved by the collaborative work of the dual structured light scanning devices on the first and second inverted six-axis robotic arms. At the same time, the entire inspection process runs automatically without manual intervention and can be applied to the inspection of steel beams of different sizes. That is, it can complete the fully automatic three-dimensional scanning and dimensional inspection of precast steel beams. Meanwhile, the computer is used to perform point cloud denoising, point cloud registration and automatic dimensional calculation. It can automatically complete the qualification judgment according to the steel structure specifications and finally output a complete inspection report.

[0034] In the description of this invention, it should be understood that the terms "upper", "lower", "bottom", "top", "front", "rear", "inner", "outer", "left", "right", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0035] While the invention has been described herein with reference to specific embodiments, it should be understood that these embodiments are merely examples of the principles and applications of the invention. Therefore, it should be understood that many modifications can be made to the exemplary embodiments, and other arrangements can be designed without departing from the spirit and scope of the invention as defined by the appended claims. It should be understood that different dependent claims and features described herein can be combined in ways different from those described in the original claims. It is also understood that features described in conjunction with individual embodiments can be used in other described embodiments.

Claims

1. A dual-robotic-arm automatic steel beam inspection system, characterized in that, include: A linear ground rail device includes two guide rails arranged opposite each other, both of which extend along the length of a steel beam. Each guide rail is also provided with a rack and a drive gear assembly that meshes with the rack. A mobile gantry assembly includes a transverse connecting beam and a movable slide block that is slidably connected to the guide rail. A vertically extending support column is mounted on the movable slide block, and both ends of the transverse connecting beam are connected to the support column. The workpiece support and positioning device includes multiple support brackets arranged at intervals along the length of the steel beam, and each support bracket is provided with a mechanical stop. Control components; Below the transverse connecting beam, there are a first inverted six-axis robotic arm and a second inverted six-axis robotic arm arranged in a spatially staggered manner. The ends of the first and second inverted six-axis robotic arms are equipped with structured light scanning devices. The control components are electrically connected to the linear ground rail device, the first inverted six-axis robotic arm, the second inverted six-axis robotic arm, and the structured light scanning device, respectively.

2. The automatic detection system for steel beams with dual robotic arms according to claim 1, characterized in that, The linear ground rail device also includes a plurality of mechanical positioning pins spaced apart along the length of the steel beam, and the movable gantry assembly has positioning elements corresponding to the mechanical positioning pins to perform segmented and precise positioning of the movable gantry assembly.

3. The automatic detection system for steel beams with dual robotic arms according to claim 1, characterized in that, The transverse connecting beam is a hollow beam structure with a cable routing channel inside.

4. The automatic detection system for steel beams with dual robotic arms according to claim 1, characterized in that, The structured light scanning device includes a structured light projector and an industrial camera. The structured light scanning device is fixed to the end effector of the robotic arm via an angle-adjustable mounting bracket.

5. A method for automatic detection of steel beams with dual robotic arms, based on the automatic detection system for steel beams with dual robotic arms according to any one of claims 1 to 4, characterized in that, Includes the following steps: Step S1: Place the steel beam on the workpiece support and positioning device and complete the initial positioning; Step S2: Use the linear ground rail device to drive the moving gantry assembly to the target position and position it using mechanical positioning pins; Step S3: Control the first inverted six-axis robotic arm and the second inverted six-axis robotic arm to move to the preset scanning pose; Step S4: Obtain point cloud data of the steel beam surface from multiple perspectives using a structured light scanning device; Step S5: Denoise and downsample the point cloud data; Step S6: Perform collaborative calibration and point cloud stitching between the first inverted six-axis robotic arm and the second inverted six-axis robotic arm to obtain a complete steel beam surface model; Step S7: Output the steel beam dimensions and weld quality inspection results.

6. The automatic detection method for steel beams with dual robotic arms according to claim 5, characterized in that, In step S4, fringe structured light is projected onto the surface of the steel beam using a structured light scanning device, and fringe deformation images are acquired using an industrial camera to calculate the three-dimensional point cloud coordinates of the steel beam surface. in, The depth value of the target point. For industrial camera focal length, This represents the baseline distance between the structured light emitter and the industrial camera. This represents the disparity value of the striped image.

7. The automatic detection method for steel beams with dual robotic arms according to claim 6, characterized in that, In step S5, the point cloud data is denoised, including: Calculate the average distance between each point in the input point cloud and its k nearest neighbors: in, For the first The average neighborhood distance of each point For the first The point and the first Euclidean distance between neighboring points; Calculate the Gaussian distribution of the average distance between all points, and then calculate the mean μ and standard deviation σ. Set distance threshold: ; Where α is the threshold coefficient; The system iterates through all points in the input point cloud. When the average neighborhood distance of a point is greater than a distance threshold, the point is identified as an outlier and deleted to achieve noise reduction of the point cloud data.

8. The automatic detection method for steel beams with dual robotic arms according to claim 7, characterized in that, In step S5, the point cloud data is downsampled, including: Based on the three-dimensional coordinate set of the point cloud data, calculate the maximum and minimum values ​​in the x, y, and z directions, and establish the minimum bounding box of the point cloud; Set the voxel mesh edge length r, and divide the three-dimensional space into voxels; Calculate the index value of each point in the voxel grid: in, , , These are the voxel indices of the point in the x, y, and z directions, respectively; Points within the same voxel mesh are grouped into a single set, and the centroid coordinates of all points within that voxel are calculated: in, The coordinates of the voxel center point; The original point cloud data within the voxel is replaced with the coordinates of the voxel center point for downsampling processing.

9. The automatic detection method for steel beams with dual robotic arms according to claim 8, characterized in that, Step S6 also includes: Step S601: Establish the spatial transformation relationship between the coordinate systems of the two robotic arms' bases: Control the first inverted six-axis robotic arm and the second inverted six-axis robotic arm to move to the preset calibration postures, so that the structured light scanning devices installed at the ends of the two robotic arms simultaneously observe a fixed common calibration target in space. The common calibration target is a planar checkerboard target or a three-dimensional spherical target. Record the pose matrices of the ends of the two robotic arms in their respective base coordinate systems. and The target pose matrix relative to the coordinate systems of the two robotic arms' ends is extracted using the target image acquired by the structured light scanning device. and Solve for the homogeneous transformation matrix between the coordinate systems of the two robotic arms' bases. ,satisfy: ； Step S602: Acquire point cloud data scanned by the two robotic arms: The first inverted six-axis robotic arm and the second inverted six-axis robotic arm scan different areas of the steel beam according to a preset trajectory, and acquire two pieces of original point cloud data through their respective structured light scanning devices. and And record the end-effector pose at each scanning moment; Step S603: Perform coarse point cloud stitching based on calibration results: using the base coordinate system transformation matrix obtained in step S601. The point cloud acquired by the first inverted six-axis robotic arm The point cloud is obtained by transforming the tool coordinate system to the base coordinate system of the second inverted six-axis robotic arm. The transformation relationship is as follows: ； This makes two point clouds and Under the same global coordinate system, preliminary spatial alignment is completed; Step S604: Optimize the point cloud stitching result using a sampling consistency initial registration algorithm based on fast point feature histograms: extract point clouds respectively. and The FPFH feature descriptor for each point in the source point cloud Several sampling points are randomly selected from the target point cloud based on feature similarity. Find the corresponding point in the matrix, construct candidate transformation matrices, and evaluate the optimal initial transformation matrix based on the distance error. To reduce the positional deviation between the two point clouds; Step S605: Use the generalized iterative nearest neighbor algorithm to perform precise registration of the point cloud.

10. The automatic detection method for steel beams with dual robotic arms according to claim 9, characterized in that, In step S605, the generalized iterative nearest neighbor algorithm is used for accurate point cloud registration, including: Given a source point cloud P and a target point cloud Q, initialize the rigid transformation matrix T; Calculate the covariance matrix in the neighborhood of each point to describe the local geometric features of the point cloud; The correspondence between the source point cloud and the target point cloud is established using the nearest neighbor search algorithm; Construct the error function: in, and These are the covariance matrices of the source point cloud and the target point cloud, respectively. Minimize the error function iteratively Continuously optimize the rigid transformation matrix This continues until the error converges, achieving accurate point cloud registration.

Images