A structure light camera based intersecting line weld detection and trajectory optimization system and method

Abstract

The application discloses a kind of based on structural light camera's intersection line weld detection and trajectory optimization system and method, including industrial computer, welding robot, welding machine and structural light camera, industrial computer includes visual processing module, motion control module and user interface module;Visual processing module is used to execute the process control of welding seam detection and welding according to the point cloud information of welding workpiece collected by the structural light camera;Motion control module is used to optimize welding seam trajectory, controls welding robot motion;User interface module provides the mode of man-machine interaction, realizes the setting of parameter, result display and report generation.Welding robot is used to drive welding torch to carry out welding operation;Welding machine is used to control the welding robot to execute welding task, and structural light camera is used to collect the image and point cloud information of welding workpiece.The application proposes detection method according to the characteristics of intersection line weld, more universal and accurate, and welding seam trajectory optimization fuses welding gun pose, which improves welding quality.

Description

Technical Field

[0001] This invention belongs to the field of welding robot technology, specifically, it relates to a system and method for detecting and optimizing the trajectory of intersecting welds based on a structured light camera. Background Technology

[0002] Intersecting welds are a commonly used welding method for multi-layer plates, offering advantages such as high joint strength, good seismic resistance, and high reliability, leading to their widespread application in aerospace, automotive, and robotics fields. However, the detection and trajectory planning of intersecting welds in tubular truss structures based on ideal mathematical models suffers from deviations during actual welding operations due to sensor noise and the influence of the welded workpiece. Therefore, intersecting weld detection and welding trajectory optimization remain challenging problems.

[0003] Traditional weld inspection and tracking methods mainly rely on manual visual inspection and measurement, which suffers from low efficiency and large errors. In recent years, with the development of 3D point cloud technology, weld inspection and tracking methods based on 3D point clouds have received increasing attention. By using equipment such as structured light cameras to acquire 3D point cloud data of the weld area, and performing preprocessing, feature extraction, and curve fitting on the point cloud data, automated weld inspection and tracking can be achieved. Compared with traditional methods, 3D point cloud-based methods have advantages such as high detection accuracy, high detection efficiency, and simple operation, which can greatly improve welding production efficiency and quality.

[0004] However, the detection and trajectory optimization of intersecting welds in tubular truss structures based on 3D point clouds still face some challenges. For example, noise and interference in the point cloud data can lead to errors and instabilities in the identification and tracking results; complex weld shapes and defects may require complex algorithms and models for processing; and deformation and changes during the welding process can also affect identification and tracking. Therefore, developing a method for intersecting weld detection and trajectory optimization based on 3D point clouds has significant research and application value.

[0005] A search revealed that no relevant patented invention currently exists that can effectively solve the above problems. For example:

[0006] Patent CN115741724A discloses a weld seam recognition and tracking system and method based on a 2D / 3D camera. The system includes an industrial control computer, a welding robot, a welding machine, a 2D vision sensor, a 3D camera, an offline programming system, and a weld seam recognition and tracking software system. The industrial control computer is used to install the weld seam recognition and tracking software system; the welding robot drives the welding torch to perform welding operations; the welding machine controls the welding robot to execute welding tasks; the offline programming system is used to plan the weld seam and control the movement of the welding robot; the weld seam recognition and tracking software system is installed on the industrial control computer and is responsible for the overall system flow control. However, this invention only provides the system design for the entire weld seam detection, without describing the specific weld seam recognition method, the specific recognition of intersecting weld seams, or the generation and optimization of welding trajectories for the generated weld seam feature points. Summary of the Invention

[0007] This invention provides a weld seam inspection and trajectory optimization system and method based on a structured light camera. It proposes a detection method and trajectory optimization method for complex weld seams and intersecting weld seams, solving the problem of low efficiency in manual visual inspection and measurement. The detection method proposed based on the inherent characteristics of intersecting weld seams is more universal and accurate. Weld seam trajectory optimization combined with welding torch pose improves welding quality.

[0008] The objective of this invention is achieved through the following technical solution: Firstly, this invention provides a system for detecting and optimizing the trajectory of intersecting weld seams based on a structured light camera. This system includes: an industrial control computer, a welding robot, a welding machine, and a structured light camera, wherein:

[0009] The industrial control computer includes a vision processing module, a motion control module, and a user interface module. The vision processing module is used to perform weld seam detection and welding process control based on the point cloud information of the welding workpiece acquired by the structured light camera. The motion control module is used to optimize the weld seam trajectory and control the movement of the welding robot. The user interface module provides a human-computer interaction method to realize parameter setting, result display, and report generation.

[0010] The welding robot is used to drive the welding torch to perform welding operations;

[0011] The welding machine is used to control the welding robot to perform welding tasks and to supply power.

[0012] The structured light camera is divided into a structured light projection module and a camera acquisition module, which are used to acquire images and point cloud information of the welded workpiece.

[0013] Furthermore, the structured light camera is mounted on one side of the welding torch using a clamp.

[0014] Furthermore, the welding torch is an automatic welding torch, and the structured light camera is fixed in the orthogonal direction of the clamp to prevent interference, and an anti-collision sensor is fixedly installed at the bottom of the welding torch holder.

[0015] Furthermore, the vision processing module in the industrial control computer includes: a 3D point cloud registration module and a 3D point cloud weld detection module;

[0016] The 3D point cloud registration module is used to perform point cloud filtering, point cloud downsampling and point cloud registration processing on the point cloud information of the welded workpiece acquired by the structured light camera, and send the processing results to the 3D point cloud weld detection module.

[0017] The 3D point cloud weld detection module is used to detect all welds and send them to the motion control module based on the cross-sectional and contour characteristics of the weld after the 3D point cloud registration module reconstructs the complete three-dimensional model of the weld.

[0018] Furthermore, the motion control module in the industrial computer includes: a weld seam trajectory optimization module and a control command module;

[0019] The weld trajectory optimization module uses the weld data detected by the 3D point cloud weld detection module, takes into account the changes in the welding torch posture, uses a four-element algorithm to interpolate the weld data, and sends the data to the control command module.

[0020] The control command module generates corresponding control commands based on the detection and trajectory optimization results to achieve automatic control during the welding process, including real-time adjustment and control of the welding torch position and welding parameters.

[0021] Furthermore, the point cloud filtering algorithms used in the 3D point cloud registration module include: pass-through filtering algorithm and statistical filtering algorithm; the point cloud downsampling methods used in the 3D point cloud registration module include: voxel method and curvature downsampling method; and the algorithm used for point cloud registration processing in the 3D point cloud registration module is the ICP algorithm.

[0022] On the other hand, this application also provides a method for detecting and optimizing the trajectory of intersecting welds based on a structured light camera, the method comprising:

[0023] Step 1: Perform point cloud registration on the point cloud data scanned by the structured light camera to obtain a 3D model;

[0024] Step 2: Based on the cross-sectional area and contour features of the intersection lines, detect all intersection line welds of the workpiece on the registered 3D model;

[0025] Step 3: After detecting all welds, optimize the weld trajectory according to the welding process requirements, so that the welding torch posture changes according to the different sections of the weld trajectory.

[0026] Further, step 1 includes:

[0027] Step 1.1: Take pictures of the workpiece from various angles using a structured light camera to obtain multi-frame point cloud data of the workpiece;

[0028] Step 1.2: Perform pass-through filtering and Gaussian filtering on the multi-frame point cloud data to remove noise points and outliers, thereby improving the point cloud quality;

[0029] Step 1.3: Perform voxel downsampling and curvature downsampling operations on the filtered point cloud data to simplify the point cloud data;

[0030] Step 1.4: Perform coarse registration of the downsampled point cloud data based on the initial matrix provided by the robotic arm;

[0031] Step 1.5: Use the ICP algorithm to perform fine registration of the point cloud based on the coarse registration result, and obtain the fine registration result for subsequent weld inspection.

[0032] Further, step 2 includes:

[0033] Step 2.1: Based on the finely registered point cloud, fit a cylinder using the iterative nearest point method;

[0034] Step 2.2: Cluster the point cloud data using the Euclidean clustering algorithm;

[0035] Step 2.3: Analyze the clustered regions, identify possible cylindrical regions, and extract the point cloud of the intersecting parts of the cylinders;

[0036] Step 2.4: Based on the characteristic that the cross-sectional area of ​​the intersection weld is elliptical and symmetrical, the point cloud of the extracted part is fitted using the least squares algorithm to obtain the complete point cloud of the intersection weld.

[0037] Furthermore, step 3 includes:

[0038] Step 3.1: Preprocess the weld trajectory data to obtain downsampled weld trajectory data;

[0039] Step 3.2: Fit the sampled weld trajectory using cubic B-splines;

[0040] Step 3.3: Use the fitted data as model points;

[0041] Step 3.4: Use the cubic B-spline algorithm and the Squad algorithm to interpolate the position and orientation of the weld trajectory to obtain the optimized weld trajectory.

[0042] The beneficial effects of this invention are:

[0043] 1. Compared with image information obtained by traditional 2D vision, the point cloud data obtained by this invention using a structured light camera is less affected by lighting conditions, and the obtained weld position information is more accurate.

[0044] 2. This invention uses a non-contact method to detect the welding area of ​​the welded workpiece, which solves the problems of inefficiency and limitations of traditional manual visual inspection.

[0045] 3. This invention addresses the problem of low recognition rate of complex curved welds due to external factors, starting from the contour characteristics of the intersection line.

[0046] 4. This invention proposes a fusion algorithm of cubic spline curve interpolation and Squad algorithm interpolation to optimize the weld trajectory. This solves the problem that most welding robots only consider the smoothness of the weld trajectory in their welding path planning and ignore the influence of the welding torch posture on the welding quality during the welding operation. Attached Figure Description

[0047] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1 This is a flowchart illustrating the detection and trajectory optimization method for intersecting weld seams.

[0049] Figure 2 This is a flowchart for inspecting intersecting weld seams.

[0050] Figure 3 A flowchart for optimizing weld seam trajectory.

[0051] Figure 4 This is a fixture diagram for a structured light camera and a welding machine.

[0052] In the diagram, 1. Anti-collision sensor, 2. Gun clamp, 3. Welding gun head, 4. Clamping part, 5. Connecting flange, 6. Structured light camera, 7. Camera connecting base. Detailed Implementation

[0053] The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0054] This application provides a system and method for detecting and optimizing the trajectory of intersecting welds based on a structured light camera, mainly used for detecting intersecting welds, and mainly includes:

[0055] The system includes an industrial control computer, a welding robot, a welding machine, and a structured light camera. The industrial control computer is responsible for the overall process control of the system. The welding robot is the actuator, responsible for driving the welding torch to perform the welding operation. The welding machine is responsible for the robot's motion execution and power supply. The structured light camera is used to acquire grayscale images and point cloud information of the workpiece to be welded.

[0056] The industrial control computer includes a vision processing module, a motion control module, and a user interface module. The vision processing module is used to perform weld seam detection and welding process control based on the point cloud information of the welding workpiece acquired by the structured light camera. The motion control module is used to optimize the weld seam trajectory and control the movement of the welding robot. The user interface module provides a human-computer interaction method to realize parameter setting, result display, and report generation.

[0057] A structured light camera acquires workpiece information and is mounted on one side of the welding torch via a fixture. It then transmits data to the industrial control computer via TCP communication. The industrial control computer acquires weld data and transmits it to the welding robot via TCP communication.

[0058] For example, the structured light camera is mounted in the orthogonal direction of the welding torch by a fixture, which avoids interference and ensures that the structured light camera has the best shooting angle. The welding torch is an automatic welding torch with sufficient space above it. The fixture is connected to the end of the robotic arm by a flange connector, and an anti-collision sensor is installed at the bottom to increase the safety of the welding process.

[0059] The vision processing module in the industrial control computer includes: a 3D point cloud registration module and a 3D point cloud weld detection module;

[0060] The 3D point cloud registration module is used to perform point cloud filtering, point cloud downsampling and point cloud registration processing on the point cloud information of the welded workpiece acquired by the structured light camera, and send the processing results to the 3D point cloud weld detection module.

[0061] The 3D point cloud weld detection module is used to detect all welds and send them to the motion control module based on the cross-sectional and contour characteristics of the weld after the 3D point cloud registration module reconstructs the complete three-dimensional model of the weld.

[0062] The 3D point cloud registration module in this embodiment mainly includes: a coarse registration module, a fine registration module, a point cloud processing module, and a data transmission module. The point cloud processing module mainly includes: point cloud filtering, point cloud downsampling, and point cloud registration. The point cloud filtering algorithms used include pass-through filtering and Gaussian filtering. Pass-through filtering is a simple and effective filtering method that removes noise and outliers from point cloud data by setting a threshold. It can quickly filter out some obvious outliers in point cloud data, but its filtering effect may not be ideal for complex point cloud data. Gaussian filtering is a smoothing filtering method based on the Gaussian function. It removes noise and outliers by smoothing the point cloud data. It can better preserve the features and structure of point cloud data and has a better filtering effect for complex point cloud data. In this invention, the two filtering methods are combined. First, pass-through filtering is used to remove obvious outliers and noise, and then Gaussian filtering is used for smoothing to obtain cleaner and smoother point cloud data. This can effectively improve the quality of point cloud data and reduce errors and uncertainties in subsequent processing. Point cloud downsampling should reduce point cloud density without losing point cloud features. This invention uses a voxel-based downsampling and curvature downsampling fusion method. First, voxel-based downsampling reduces the quantity and density of point cloud data to an appropriate level. Then, curvature downsampling is used for further sampling and filtering to obtain more suitable point cloud data. This effectively improves the efficiency of point cloud data processing and analysis, reduces computation time and storage space, while preserving the main features and shape of the point cloud data. The point cloud registration algorithm used is an auxiliary registration algorithm combined with robotic arm assistance. Coarse registration is performed based on the initial transformation matrix provided by the robotic arm and the hand-eye matrix after hand-eye calibration. Fine registration is then performed on the coarsely registered point cloud using the ICP algorithm to obtain the finely registered result.

[0063] Figure 1 This is a schematic diagram illustrating the process of intersection weld detection and trajectory optimization based on a structured light camera, provided in a specific embodiment of the present invention. For example... Figure 1 As shown, the intrinsic parameters of the structured light camera are first calibrated. After calibration, the structured light camera is mounted on one side of the welding machine using a fixture, and then mounted on the end effector of the welding robot. Its position relative to the robot is then calibrated. After visual calibration, the robot is moved to a suitable angle to acquire multiple frames of point cloud data. Point cloud preprocessing is performed on each frame, including point cloud filtering and downsampling. Finally, simplified point cloud data retaining the main features and shape is obtained. Based on the hand-eye matrix obtained from hand-eye calibration and the initial pose point cloud provided by the robot, coarse registration is performed. Fine registration of the point cloud is then completed using the ICP algorithm. If the point cloud registration is successful, Euclidean clustering is performed on the registered point cloud to obtain different clustering regions.

[0064] The motion control module in the industrial computer includes: an intersection weld detection module, a weld trajectory optimization module, and a control command module;

[0065] The intersection weld detection module first divides the clustered region into cylinders, then extracts the point cloud of the welding area based on the cross-sectional area and contour characteristics of the intersection weld, and finally fits the intersection weld and sends it to the weld trajectory planning module.

[0066] The weld trajectory optimization module uses the weld data detected by the intersection line weld detection module, takes into account the changes in the welding torch posture, uses a four-element algorithm to interpolate the data, and sends the data to the control command module.

[0067] The control command module generates corresponding control commands based on the detection and trajectory optimization results to achieve automatic control during the welding process, including real-time adjustment and control of the welding torch position and welding parameters.

[0068] The weld trajectory planning module receives the weld data, first optimizes the weld trajectory, and takes into account the influence of the welding torch posture on the welding quality during the welding process. It uses the cubic B-spline algorithm and the Squad algorithm to interpolate the position and posture of the weld trajectory, fuses the welding torch pose with the weld trajectory data, optimizes the fused data, and finally sends the optimized welding path plan to the industrial control computer to execute the final welding operation.

[0069] Figure 2 This is a schematic diagram of the intersecting weld inspection process provided in a specific embodiment of the present invention. For example... Figure 2 As shown, considering the large curvature variation in the welding area of ​​the intersecting weld, a curvature downsampling method is used to retain the main features of the welding area while removing some redundant point clouds. Cluster analysis is performed on the downsampled point clouds, and multiple cylinders are fitted to the clustered regions using the RANSAC algorithm. The point cloud set Q = {q1, q2, ... q...} of the intersection lines of the cylinders is then extracted. n Considering that the coordinate system of the point cloud has been transformed to the world coordinate system during ICP registration, the truss structural components are placed on a plane parallel to the XYO plane of the world coordinate system. The extracted point cloud is projected onto the XYO plane. Given that the cross-section of the intersection line is elliptical or circular and has inflection points, the least squares method is used to fit the ellipse equation. The ellipse equation fitted to the detected n points is as follows:

[0070]

[0071] Where A, B, C, D, and E are the coefficients of the ellipse fitted by the least squares method.

[0072] Let the set of elliptical points obtained by plane fitting be P = {p1, p2, ..., p...} n}, iterate through the distances from point set P to point set Q, and take the minimum distance d. i Place it into array D. i The calculation formula is as follows:

[0073]

[0074] Where p i (x i ,y i ), q j (x j ,y j ,z j ), p i The points q are in the set of elliptical points obtained by plane fitting. i It is the point of the cloud of intersecting points Q.

[0075] If there are multiple points q j With p i distance d i If the value is less than the threshold of 0.3 mm, select the smallest z. min point q min As a new key point of weld, z min It is multiple points q j The minimum value of the Z-coordinate, if there exists only one q. j With p i distance d i If it is less than the threshold of 0.3mm, then q j This is considered a new critical weld point, i.e.

[0076] q min (x min ,y min ,z min )

[0077]

[0078] o n (x n ,y n ,z n )

[0079] The point cloud portion is extracted by inversely calculating the fitted ellipse equation. n As the key point of the intersecting weld, it is placed in the point set Q and sent to the weld trajectory optimization module.

[0080] Considering the special characteristics of the intersection weld, the crescent-shaped oscillating welding method is adopted. The calculation formulas for the welding speed and wire feed rate are as follows:

[0081]

[0082]

[0083] Where V is the welding speed, d is the wire radius, I is the current, U is the voltage, β is the wire feed rate, H is the oscillation frequency, and L is the oscillation length.

[0084] Figure 3 This is a schematic diagram of the weld trajectory optimization module provided in a specific embodiment of the present invention. Figure 3 As shown, in the weld trajectory data, the weld contour point cloud data is extracted using equal-interval sampling to achieve sparse operation on the weld trajectory. The key points of the fitted intersection line are used as sampling objects, and sampling is performed at Euclidean spatial distances of interval d to extract the weld contour point cloud data where the scanned feature points of the weld are located. The selection of the Euclidean distance d should be based on the curvature of the weld trajectory so that the sampled weld trajectory can better reflect the shape of the weld trajectory. To reduce sampling errors, a cubic B-spline algorithm is used to fit the sampled weld trajectory data. Then, using the fitted data points as shape points, the position and orientation of the weld trajectory are interpolated using both the cubic B-spline algorithm and the Squad algorithm. Finally, a smooth welding trajectory is obtained.

[0085] The mounting fixture is shown in the diagram. Figure 4 The clamping part 4 is connected to the welding torch head 3 by a thread. The torch clamp 2 is used to clamp the welding torch. The torch clamp 2 is connected to the camera connecting base 7 by bolts. The structure light camera 6 is connected to the camera connecting base 7 by bolts. The anti-collision sensor 1 is installed on the torch clamp 2. The connecting flange 5 is fixed to the robotic arm and the torch clamp 2 by bolts.

[0086] For example, the patent provides a method for detecting and optimizing the trajectory of intersecting welds and V-groove welds using the above-mentioned structured light camera-based system, including the following specific steps:

[0087] S1: Hand-eye calibration and TCP calibration of the robotic arm; a structured light camera captures the 3D point cloud of the welding workpiece; the industrial control computer loads the model point cloud of the workpiece; and the point cloud registration method is used for precise positioning.

[0088] In this embodiment, step S1 includes:

[0089] Step S11: Robotic arm hand-eye calibration and TCP calibration;

[0090] Step S12: The structured light camera module captures the point cloud of the workpiece;

[0091] Step S13: Perform point cloud registration on the two point clouds to obtain the rotation and translation matrix between the two point clouds;

[0092] Step S14: Obtain the rotation and translation matrix from S13, i.e., the fine positioning result, through offline programming;

[0093] Step S15: Perform 3D reconstruction on the registered point cloud.

[0094] In this embodiment, step S2 includes:

[0095] Step S21: Use the workpiece registration point cloud obtained in step S15 of step S1;

[0096] Step S22: Select the weld type to be inspected and simplify the point cloud using point cloud preprocessing;

[0097] Step S23: The simplified point cloud is used to detect the intersecting welds in the model using a weld recognition algorithm;

[0098] Step S24: Transmit the identification results from S22 to the robot via TCP communication for welding.

[0099] In this embodiment, step S3 includes:

[0100] Step S31: Combine the weld seam trajectory in S23 with the welding torch posture for data fusion and further optimization;

[0101] Step S32: Determine the initial point based on the optimized weld trajectory;

[0102] Step S33: Start soldering from the initial solder joint.

[0103] The above embodiments are used to explain and illustrate the present invention, but not to limit the present invention. Any modifications and changes made to the present invention within the spirit and scope of the claims shall fall within the protection scope of the present invention.

Claims

1. A system for detecting and optimizing the trajectory of intersecting welds based on a structured light camera, characterized in that, The system includes: an industrial control computer, a welding robot, a welding machine, and a structured light camera, among which: The industrial control computer includes a vision processing module, a motion control module, and a user interface module. The vision processing module is used to perform weld seam detection and welding process control based on the point cloud information of the welding workpiece acquired by the structured light camera. The motion control module is used to optimize the weld seam trajectory and control the movement of the welding robot. The user interface module provides a human-computer interaction method to realize parameter setting, result display, and report generation. The vision processing module in the industrial control computer includes: a 3D point cloud registration module and a 3D point cloud weld detection module; The 3D point cloud registration module is used to perform point cloud filtering, point cloud downsampling and point cloud registration processing on the point cloud information of the welded workpiece acquired by the structured light camera, and send the processing results to the 3D point cloud weld detection module. The 3D point cloud weld detection module is used to detect all welds and send them to the motion control module based on the cross-sectional and contour characteristics of the weld after the 3D point cloud registration module reconstructs the complete three-dimensional model of the weld. The motion control module in the industrial computer includes: a weld seam trajectory optimization module and a control command module; The weld trajectory optimization module uses the weld data detected by the 3D point cloud weld detection module, takes into account the changes in the welding torch posture, uses a four-element algorithm to interpolate the weld data, and sends the data to the control command module. The control command module generates corresponding control commands based on the detection and trajectory optimization results to achieve automatic control during the welding process, including real-time adjustment and control of the welding torch position and welding parameters. The welding robot is used to drive the welding torch to perform welding operations; The welding machine is used to control the welding robot to perform welding tasks and to supply power. The structured light camera is divided into a structured light projection module and a camera acquisition module, which are used to acquire images and point cloud information of the welded workpiece.

2. The system for detecting and optimizing the trajectory of intersecting welds based on a structured light camera according to claim 1, characterized in that, The structured light camera is mounted on one side of the welding torch using a clamp.

3. The system for detecting and optimizing the trajectory of intersecting welds based on a structured light camera according to claim 1, characterized in that, The welding torch is an automatic welding torch, and the structured light camera is fixed in the orthogonal direction of the clamp to prevent interference. An anti-collision sensor is fixedly installed at the bottom of the welding torch holder.

4. The system for detecting and optimizing the trajectory of intersecting welds based on a structured light camera according to claim 1, characterized in that, The point cloud filtering algorithms used in the 3D point cloud registration module include: pass-through filtering algorithm and statistical filtering algorithm; the point cloud downsampling methods used in the 3D point cloud registration module include: voxel method and curvature downsampling method; the algorithm used for point cloud registration processing in the 3D point cloud registration module is ICP algorithm.

5. A method for detecting and optimizing the trajectory of intersecting welds based on a structured light camera, based on the intersecting weld detection and trajectory optimization system according to any one of claims 1-4, characterized in that, The method includes: Step 1: Perform point cloud registration on the point cloud data scanned by the structured light camera to obtain a 3D model; specifically including: Step 1.1: Take pictures of the workpiece from various angles using a structured light camera to obtain multi-frame point cloud data of the workpiece; Step 1.2: Perform pass-through filtering and Gaussian filtering on the multi-frame point cloud data to remove noise points and outliers, thereby improving the point cloud quality; Step 1.3: Perform voxel downsampling and curvature downsampling operations on the filtered point cloud data to simplify the point cloud data; Step 1.4: Perform coarse registration of the downsampled point cloud data based on the initial matrix provided by the robotic arm; Step 1.5: Use the ICP algorithm to perform fine registration of the point cloud coarse registration result to obtain the fine registration result, which will be used for subsequent weld inspection; Step 2: Based on the cross-sectional area and contour features of the intersection lines, detect all intersection line welds of the workpiece using the registered 3D model; specifically including: Step 2.1: Based on the finely registered point cloud, fit a cylinder using the iterative nearest point method; Step 2.2: Cluster the point cloud data using the Euclidean clustering algorithm; Step 2.3: Analyze the clustered regions, identify possible cylindrical regions, and extract the point cloud of the intersecting parts of the cylinders; Step 2.4: Based on the characteristic that the cross-sectional area of ​​the intersection weld is elliptical and symmetrical, the point cloud of the extracted part is fitted using the least squares algorithm to obtain the complete point cloud at the intersection weld. Step 3: After detecting all welds, optimize the weld trajectory according to the welding process requirements, so that the welding torch posture changes according to the different sections of the weld trajectory; specifically including: Step 3.1: Preprocess the weld trajectory data to obtain downsampled weld trajectory data; Step 3.2: Fit the sampled weld trajectory using cubic B-splines; Step 3.3: Use the fitted data as model points; Step 3.4: Use the cubic B-spline algorithm and the Squad algorithm to interpolate the position and orientation of the weld trajectory to obtain the optimized weld trajectory.

Images