Box type workpiece welding seam positioning method, system and robot based on three-dimensional point cloud

Abstract

The invention relates to the technical field of intelligent welding, and provides a box type workpiece welding seam positioning method, system and robot based on a three-dimensional point cloud.The method comprises the steps that the three-dimensional point cloud of a box type target welding seam area is obtained, and the welding seam path reference direction is recognized; performing plane detection and fitting on the point cloud of the welding seam area, and performing intelligent plane screening based on an angle threshold value of an included angle between the reference direction and a plane normal vector to obtain a plurality of candidate planes; constructing a comprehensive evaluation function F including an included angle score, a distance score, a flatness score and a thickness score, and screening the candidate planes to obtain a target plane; and then the target plane is biased in the normal vector direction to generate a plane family, the plane family intersects with the triangular net curved surface of the welding area, a welding seam intersecting line is formed, and a track point column used for welding is obtained. According to the method, high-precision automatic identification and trajectory planning of the welding seam path of the complex structure are achieved without manual teaching, and the intelligent level of the welding process and the welding seam identification reliability are improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent welding technology, specifically to a method, system, and robot for locating weld seams on box-type workpieces based on three-dimensional point clouds. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] In nuclear power plant construction, the ring lifting beam, used to hoist important equipment, is a critical load-bearing component, and its safety directly affects the operational stability of the entire project. The ring lifting beam is typically composed of multiple box-shaped structures spliced ​​together to form a complete ring segment. These boxes are connected by welding, and each standard ring segment usually requires eight welds, including welds in the vertical direction and on the curved bottom surface. Due to the extremely high safety requirements of nuclear power equipment, the quality of these welds must be highly reliable; therefore, the precision and consistency of the welding path must be ensured during the welding process.

[0004] Currently, in the welding operation of ring lifting beams, a horizontal assembly method is usually adopted, that is, the components are laid flat before welding. Since the bottom surface of these components is a curved arc, the welding path is located on a complex curved surface. In addition, the surface is reflective after grinding, and the clamping position may be inaccurate. At the same time, there are many external interference factors such as ambient light. Traditional methods are difficult to achieve accurate and stable weld identification and welding path planning. Most existing methods rely on manual "teaching" (manually guiding the robot to weld each time) or recognition technology based on two-dimensional images. These methods have the following problems: (I) Difficulty in identifying welding feature surfaces: The bottom of the ring lifting beam is a curved surface, and the weld position is distributed in multiple different directions and planes in space, resulting in diverse and irregular geometric features of the welding area. Two-dimensional images can only reflect planar information and cannot accurately restore these three-dimensional spatial features, causing identification deviation. (II) Unstable extraction of weld toe position: The weld toe refers to the edge point where the weld and the base material meet, that is, the start / end point of the weld or the boundary line of the welding trajectory, which is usually located near the junction of two components. Because the surface is polished before welding, the reflection is strong, making it difficult for the vision system to obtain a clear boundary. In addition, the welding gap is small and the color contrast is not obvious, resulting in low accuracy and poor repeatability of automatic extraction of weld toes. Summary of the Invention

[0005] To address the aforementioned problems, this invention proposes a method, system, and robot for locating weld seams in box-shaped workpieces based on three-dimensional point clouds. It introduces a weld seam recognition and trajectory planning method based on three-dimensional point clouds, and combines a multi-dimensional comprehensive evaluation function and a geometric intersection strategy to achieve automatic generation of weld seam paths without teaching under complex spatial structures, significantly improving welding accuracy and intelligence.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: One or more embodiments provide a method for locating weld seams on box-type workpieces based on three-dimensional point clouds, including the following steps: The three-dimensional point cloud of the weld seam area of ​​the target to be welded in the box-type object is obtained. The path direction vector is calculated based on the centroid of the point cloud of adjacent scanning positions and normalized to obtain the reference direction of the weld seam path. Plane detection and fitting are performed based on point cloud data of the target weld area to obtain the fitted plane; intelligent plane filtering is performed based on the angle threshold between the weld path reference direction and the plane normal vector to obtain multiple candidate planes. A comprehensive evaluation function F, including angle score, distance score, flatness score, and thickness score, is constructed to filter multiple candidate planes and obtain the target plane. The target plane is used to generate a family of planes by offsetting the direction of the normal vector, and then geometrically intersected with the triangular mesh surface of the welding target area to form the weld intersection line; The weld intersection line is projected back onto the target plane and resampled to form the final trajectory point sequence used for welding, which can be used for robotic automated welding.

[0007] One or more embodiments provide a weld seam positioning system for box-type workpieces based on three-dimensional point clouds, including: The weld path reference direction recognition module is configured to acquire the three-dimensional point cloud of the weld area of ​​the target to be welded in the box-type class, calculate the path direction vector based on the centroid of the point cloud of adjacent scanning positions and normalize it to obtain the weld path reference direction. The local plane fitting module is configured to perform plane detection and fitting based on the point cloud data of the target weld area to obtain the fitted plane; and to perform intelligent plane filtering based on the angle threshold between the weld path reference direction and the plane normal vector to obtain multiple candidate planes. The plane selection module is configured to construct a comprehensive evaluation function F, which includes angle score, distance score, flatness score, and thickness score, to filter multiple candidate planes and obtain the target plane; The geometric intersection module is configured to generate a family of planes from the target plane by offsetting the direction of the normal vector, and then perform geometric intersection with the triangular mesh surface of the welding target area to form the weld intersection line; The projection module is configured to project the weld intersection line back to the target plane and resample it to form the final trajectory point sequence used for welding.

[0008] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention addresses the difficulty of identifying weld feature surfaces by utilizing the spatial geometric reconstruction capabilities of 3D point clouds to fully restore the 3D morphology of the weld path, overcoming the shortcomings of traditional 2D images in weld geometry reconstruction and significantly improving the accuracy and stability of path recognition. To address the issue of insufficient robustness in weld recognition, a multi-dimensional scoring mechanism is employed to filter weld target planes, enhancing the system's adaptability to complex welding scenarios. To address the problem of unstable weld toe position extraction, a geometric intersection and resampling strategy is used to stably extract weld boundaries, ensuring the continuous usability of the trajectory and guaranteeing the repeatability and controllability of the welding path. This method is particularly suitable for welding operations in critical structures such as nuclear power plant ring beams, significantly improving the practicality and reliability of automated welding systems in high-safety scenarios. The advantages and additional benefits of this invention will be described in detail in the following specific embodiments. Attached Figure Description

[0009] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute a limitation thereof.

[0010] Figure 1 This is a flowchart of the teachingless welding trajectory planning method of Embodiment 1 of the present invention; Figure 2 This is a schematic diagram of the ring lifting beam in Embodiment 1 of the present invention. Detailed Implementation

[0011] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0012] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0013] It should be noted that the terminology used herein is for describing particular embodiments only and is not intended to limit the exemplary embodiments of the present invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof. It should be noted that, without conflict, the various embodiments and features within those embodiments can be combined with each other. The embodiments will now be described in detail with reference to the accompanying drawings.

[0014] Example 1 In one or more of the technical solutions disclosed in the implementation methods, such as Figures 1 to 2As shown, a method for locating weld seams in box-shaped workpieces based on 3D point clouds includes the following steps: Step 1: Obtain the 3D point cloud of the weld seam area of ​​the target object to be welded in the box-type class. Calculate the path direction vector based on the centroid of the point cloud of adjacent scanning positions and normalize it to obtain the reference direction of the weld seam path. Step 2: Perform plane detection and fitting based on the point cloud data of the target weld area to obtain the fitted plane; perform intelligent plane filtering based on the angle threshold between the weld path reference direction and the plane normal vector to obtain multiple candidate planes; Step 3: Construct a comprehensive evaluation function F that includes angle score, distance score, flatness score, and thickness score to filter multiple candidate planes and obtain the target plane; Step 4: Generate a family of planes from the target plane by offsetting the direction of the normal vector, and geometrically intersect them with the triangular mesh surface of the welding target area to form the weld intersection line; Step 5: Project the weld intersection line back to the target plane and resample to form the final trajectory point sequence used for welding; This implementation acquires 3D point cloud information of the area to be welded in a box-like structure. The reference direction of the weld path is calculated using the spatial distribution between the centroids of the point cloud, ensuring the path direction aligns with the actual weld orientation. Subsequently, plane fitting is performed based on the point cloud. By setting a threshold for the angle between the path reference direction and the plane normal vector, candidate planes meeting geometric characteristics are initially selected. Next, a comprehensive evaluation function F is constructed to score the candidate planes across four dimensions: angle matching degree, distance from the path, plane flatness, and surface thickness consistency, to select the optimal target plane. After normal offsetting, the target plane forms a family of planes, which are used to intersect with the target triangular mesh surface geometry to obtain the precise weld intersection line. Finally, the intersection line is projected onto the target plane and resampled to form a welding trajectory point sequence, guiding automated welding equipment to complete the weld welding operation.

[0015] This implementation addresses the challenge of identifying weld feature surfaces by leveraging the spatial geometric reconstruction capabilities of 3D point clouds to fully restore the 3D morphology of the weld path. This overcomes the limitations of traditional 2D images in weld geometry reconstruction, significantly improving the accuracy and stability of path recognition. To address the issue of insufficient robustness in weld recognition, a multi-dimensional scoring mechanism is employed to filter weld target planes, enhancing the system's adaptability to complex welding scenarios. To address the instability in weld toe position extraction, a geometric intersection and resampling strategy is used to stably extract weld boundaries, ensuring continuous usability of the trajectory and guaranteeing the repeatability and controllability of the welding path. This method is particularly suitable for welding operations in critical structures such as nuclear power plant ring beams, significantly improving the practicality and reliability of automated welding systems in high-safety scenarios.

[0016] The above steps in this embodiment are the processing steps after acquiring a three-dimensional point cloud for a weld. Multiple angles and positions are scanned for a weld to obtain a three-dimensional point cloud for the same weld. Preferably, two scan images of the start and end points are acquired for each weld area to obtain three-dimensional point cloud data. For example, for a box body of a ring beam with 8 welds, steps 1 to 5 above need to be performed for each weld to identify the weld. Then the next weld is scanned and identified through the above steps until all welds are identified. Step 1 involves acquiring the 3D point cloud of the target weld area, calculating the path direction vector based on the centroids of adjacent scanned point clouds, and normalizing the vectors to obtain the weld path reference direction. This method includes the following steps: Step 11: Scan the target to be welded from different scanning positions to obtain three-dimensional point cloud data and perform preprocessing; The 3D point cloud data collected for ring-type girder box-shaped components includes multiple point cloud files, each corresponding to a scanned position of the box-shaped component.

[0017] Specifically, for each weld position of each box-type component, the scanner will perform multiple scans at different locations; each scan generates a point cloud file to record the three-dimensional shape from the corresponding viewpoint. Multiple such files are combined to form a complete box-type component model. like Figure 2 As shown, the scanning method is as follows: Before the automatic welding begins, each part of the box is fixed in the corresponding position so that the gaps on the component plate meet the welding requirements; for example, for a transverse weld, a robot or other welding execution equipment carries a sensor to a designated position to collect the point cloud of the workpiece surface, and changes the scanning position to continue collecting the point cloud until the point cloud collection of all positions of the current weld is completed.

[0018] Furthermore, methods for preprocessing the obtained 3D point cloud include: Step 111: Filter the point cloud data along the Z-axis to remove points that exceed the set range, in order to ensure the validity of the point cloud data. For each point cloud data, check the Z-axis (height) coordinates of all points and remove outliers that exceed the set range. Outliers include ground noise, supports, and reflective interference. Step 112: Perform voxelization downsampling on the filtered point cloud data, dividing the point cloud data space into multiple cubic grids, i.e., voxels. Each grid retains one representative point to reduce the number of points without affecting the overall shape. The voxelization downsampling in this embodiment can reduce the amount of point cloud data, improve processing efficiency, and at the same time maintain the geometric features of box-type components.

[0019] Step 12, Intelligent Path Direction Analysis: For point cloud data at different scanning positions, calculate the centroid coordinates of the point cloud at each scanning position, construct the weld path direction vector from the centroids of the point cloud data obtained from two adjacent frames of scanning, and normalize it. Furthermore, in this embodiment, the weld seams are arranged according to the direction of the coordinate axis, and corresponding fixtures are set so that the welding object and the robot have a fixed relative position; When the coordinate system of the welding box is in the same coordinate system as the welding robot, the weld seam is placed according to the axis of the coordinate system. The weld seam on the lower surface of the box is in the X-axis and Y-axis direction, and the vertical weld seam is in the Z-axis direction. Since the point cloud data being processed is scanned for a single weld seam, the point cloud images at different scanning positions before and after can cover the weld seam, thus allowing for a rough identification of the weld seam's direction. Specifically, the centroids of the point cloud data obtained from two consecutive scan frames are selected to construct the weld path direction vector, using the following formula: ; in, and These are the centroid coordinates of the point cloud data for the previous and next frames, respectively.

[0020] Specifically, the path direction vector is normalized to obtain the unit path vector: ; in, These represent the unit components of the path direction on the X, Y, and Z axes of the world coordinate system, respectively.

[0021] Step 13: Take the direction corresponding to the largest unit component of the normalized unit path vector as the reference direction for the weld path. Specifically, the dominant direction is determined by comparing the projections of the path direction onto each coordinate axis, and the unit vector of this direction is defined as the reference direction for the weld path. , making ,in , , That is, the dominant direction is the X-axis, Y-axis, or Z-axis direction; Since the curvature of the curved surface is affected by the actual processing and its curvature and other information are uncontrollable, this embodiment extracts a relatively controllable vertical plane A during the weld positioning process. This plane is then translated a specified distance along its own normal vector to obtain plane A'. The intersection of plane A' and the curved surface point cloud is calculated, and the intersection line projected back onto plane A is the weld. Plane detection is then performed. In step 2, planar detection and fitting are performed based on the point cloud data of the target weld area. The point cloud data is processed by introducing the maximum connected component extraction (MQC) algorithm to achieve multi-plane detection and fitting, including the following steps: Step 21: Perform initial plane fitting through random sampling to obtain the initial plane; Randomly select 3 non-collinear points from the point cloud. Calculate the equation of the plane determined by these three points; Calculate two sets of vectors representing the two edges: ; ; Plane normal vector Calculated through vector cross product ; Normalized normal vector: ; The equation of the initial plane is: ; Where a, b, and c are the normal vector components of the plane, and d is the constant term in the plane equation; Coordinates of the known point Substituting the equation of the initial plane, calculate the plane equation coefficients d: ; Step 22: Based on the consistency between geometric distance and normal, filter out the candidate set of interior points of the initial plane. Interior points are points that belong to the plane. For each point in the point cloud Calculate its distance to the plane: ; Distance threshold judgment: If (System settings) If the point is an interior point candidate, then that point is considered an interior point candidate.

[0022] Optionally, it also includes direction consistency judgment and verification of the normal of the point. With plane normal The included angle: ; if If the condition is met, the point is accepted as an interior point candidate; all interior point candidates are collected to form an interior point candidate set. ; This embodiment sets a threshold. ; Step 23, Maximum Connected Component Extraction: Candidate Set of Interior Points Perform connectivity analysis and retain the largest connected component as the final set of interior points; First, a spatial index structure (KD-tree) is constructed for the candidate set of interior points, and then Euclidean clustering algorithm is used for connectivity analysis.

[0023] Step 24, Plane Fitting Optimization: Based on the number of points in the maximum connected component, identify the effective set of points in the plane, and use the least squares method to fit the plane iteratively based on the RANSAC algorithm to obtain the fitted plane; If the number of points in the largest connected component is greater than the set threshold Nmin, then the current planar model is valid. Record the current best model, the planar model with the largest number of points in the largest connected component. The formula for calculating the RANSAC algorithm iteration count I is: ; in, For confidence level, For interior point proportions, This represents the number of sampling points per time.

[0024] For the finally selected planar model, a new planar fit is performed using all points in the largest connected component. The planar fit is performed using the least squares method, and the optimized planar parameters (a, b, c, d) are obtained by solving for them.

[0025] When the maximum number of iterations is reached The iteration terminates when the number of interior points reaches the theoretical maximum. Output the plane model with the highest number of connected component points and its optimized parameters. Remove detected interior points from the remaining point cloud and continue detecting the next plane until the remaining point cloud is insufficient to form a plane.

[0026] RANSAC (Random Sample Consensus Algorithm) is an iterative method for estimating the parameters of a mathematical model from a dataset containing outliers. Under set distance thresholds and maximum iteration counts, an improved RANSAC algorithm is used for multi-plane detection and fitting of point clouds. The improved RANSAC algorithm, by incorporating connectivity analysis into the detection process, effectively optimizes interior point statistics and model evaluation, retaining only the largest connected component as valid interior points and removing isolated small regions, thus ensuring the integrity and continuity of the plane. This is particularly suitable for box-type parts, whose planes are typically large, continuous areas, and welds are usually located on continuous planes. The optimized planes ensure the continuity of the weld path, improving the accuracy of subsequent path planning and welding operations. This optimization is an improvement on the RANSAC detection process, enhancing the quality and reliability of plane detection through connectivity constraints.

[0027] Furthermore, each plane fitting result is parameterized and encapsulated in a plane data container A for use in subsequent steps such as plane selection, intersection calculation, and feature matching.

[0028] From the multiple planes obtained from the initial fitting, qualified welding surfaces that may contain welds are selected, and surfaces that are too small, too skewed, or invalid are removed to ensure the accuracy of subsequent weld extraction.

[0029] In step 2, intelligent plane selection: Based on an angle threshold between the weld path reference direction and the plane normal vector, intelligent plane selection is performed to obtain multiple candidate planes. This method includes the following steps: Step 201: Filtering based on plane size: Calculate the extreme value difference of the plane point cloud in the X, Y, and Z directions according to the preset plane width threshold, and filter out the planes whose extreme value difference in all three directions is lower than the threshold. Specifically, the extreme value difference is the difference between the maximum and minimum values ​​of each plane in the X, Y, and Z directions, i.e., the range. If the difference in all three directions is lower than the set width threshold, it means that the plane is too small and is filtered out.

[0030] Step 202, Filtering based on direction and angle: Set angle threshold For each weld path direction vector Filter out plane normal vectors With weld path direction vector Planes with an included angle greater than a set angle threshold are selected as candidate planes.

[0031] In this embodiment, in the plane obtained by fitting in the previous step, if the surface corresponding to the weld is a curved arc surface, the plane fitting will divide the arc surface into multiple small planes, i.e. approximate planes. In this way, the angle between the area close to the weld and the weld is close to 0 degrees, and the plane far from the weld will have a certain angle with the weld. By filtering by angle threshold, the plane closest to the weld can be found, thereby identifying the plane where the weld is located as a candidate plane.

[0032] Calculate the normal vector for each candidate plane. With path direction vector The included angle: ; Set the included angle threshold ,choose The plane is used as the candidate target plane.

[0033] Based on the path direction selection: For the weld path direction vector in the X direction, the plane with the smaller angle between the normal vector and the X-axis is preferred; for the weld path direction vector in the Y direction, the plane with the smaller angle between the normal vector and the Y-axis is preferred; for the weld path direction vector in the Y direction, the plane with the smaller angle between the normal vector and the Z-axis is preferred.

[0034] Step 3, Multi-objective Optimization Selection: When multiple candidate planes exist, a comprehensive evaluation function F is established, which includes angle score, distance score, flatness score, and thickness score. The formula is as follows: ; in, , , , These are weighting coefficients, corresponding to the weights of the angle score, distance score, flatness score, and thickness, respectively, satisfying... The weight values ​​are set according to the process requirements; The included angle score reflects the consistency between the plane normal vector and the weld path direction; The distance score reflects the distance from the plane to the outer surface of the box-shaped part; the closer the distance, the higher the score. The flatness score reflects the smoothness of the surface of the plane; the higher the flatness, the higher the score. For thickness scoring, this embodiment uses surface point cloud data, which can be set to 0; The method for setting the weighting coefficients is as follows: Angle weight w1: When the accuracy requirement of the weld direction is high, increase w1; Distance weight w2: Increase w2 when it is necessary to avoid other interfering surfaces; Flatness weight w3: Increase w3 when high surface quality is required; Preferably, in this embodiment, w1 = 0.35, w2 = 0.1, and w3 = 0.55; Choose the plane with the largest F-value as the final target plane. .

[0035] In step 4, the welding trajectory is generated by offsetting the normal vector direction of the selected planes to create a family of planes. This family of planes is then geometrically intersected with the triangular mesh surface of the welding target area to form the weld intersection line. The method includes the following steps: Step 41, Surface Construction: Smooth the point cloud of the target weld area, and reconstruct the three-dimensional discrete point cloud of the weld area into a continuous surface S using a triangular mesh method; Step 42, Normal translation plane cutting: along the target plane normal vector Make a certain offset This yields a series of parallel virtual cutting planes, resulting in a family of planes. Normal vector along the target plane For the target plane By applying bias, a planar family is obtained. The formula is: ; Step 43, Geometric Intersection: Calculate the family of planes The intersection line with the continuous surface S is obtained as a polygonal line; after multiple samplings, only the polygonal lines that satisfy the weld zone constraint are retained as the weld intersection lines.

[0036] Specifically, weld seam constraints refer to the geometric and technological constraints used to filter the polygonal lines obtained from the intersection of a plane and a curved surface. These constraints are used to select the weld seam intersection line that meets the welding process requirements from multiple polygonal lines obtained through multiple sampling. Specifically, they include the following constraints: Directional constraints: The main extension direction of the intersection line should be consistent with the weld path direction. Calculate the main direction vector of the intersection line, and the angle between it and the weld path direction vector v should be less than the preset angle threshold to ensure that the direction of the intersection line is consistent with the welding process requirements. Boundary constraints: The intersection line should be within the usable range and should not conflict with the workpiece boundary, internal structure or functional area. Check whether all points of the intersection line are within the effective welding area of ​​the workpiece to avoid conflicts with the boundary, holes, internal structure, etc.

[0037] Continuity constraints: The intersection line should have good continuity, avoiding obvious breaks or jumps. Check the distance between adjacent points of the intersection line to ensure that the distance is within a reasonable range and avoid abnormally large gaps. Ensure that the intersection line can be used as a continuous welding path. During multiple offset sampling processes, the constraints mentioned above are checked sequentially for each intersection polygonal line. Only when a polygonal line simultaneously satisfies all constraints is it considered a candidate weld intersection line.

[0038] The line with the highest overall score was ultimately selected as the weld intersection line for subsequent trajectory generation and welding path planning.

[0039] In step 5, the points on the weld intersection line are projected onto the target plane one by one. The final set of weld trajectory points is obtained, the projected points are sorted according to connectivity, and smoothing and equidistant resampling are performed to obtain the weld trajectory point list. The points on the weld intersection line are projected onto the target plane one by one. This can be done by projecting along the plane normal vector, keeping the overall shape of the point cloud unchanged. Specifically, in this embodiment, the points are translated along the plane normal vector.

[0040] This step, through processing steps such as surface reconstruction, plane cutting, intersection extraction, geometric projection, and trajectory resampling, enables the automatic generation of welding paths that meet process constraints on the surface of complex 3D components, providing a high-precision trajectory point list that can be directly called upon for subsequent intelligent welding.

[0041] The above process completes the automatic identification of a weld seam. Then, the above process is repeated to scan the next weld seam, automatically identify the weld seam and weld it, until the welding of one box is completed, and then the process moves to the next box for welding. Furthermore, multi-plane coordination is employed during the welding process, including welding sequence planning: priority is given to welding seams along the X and Y axes, and a symmetrical welding strategy is used to reduce welding deformation. After completing the XY plane welds, vertical welding along the Z direction is then performed.

[0042] Path coordination and conflict avoidance: Coordinate weld paths in different directions to ensure continuous and interference-free welding torch movement. Optimize the path according to the welding sequence to improve efficiency, ensure weld quality, and reduce thermal deformation and residual stress.

[0043] Example 2 Based on Embodiment 1, this embodiment provides a robot configured to perform the three-dimensional point cloud-based box-type workpiece weld seam positioning method described in Embodiment 1 to identify and locate the weld seam position and weld the object to be welded; it also includes a fixing device disposed opposite to the robot, the fixing device being used to fix the object to be welded.

[0044] Example 3 Based on Embodiment 1, this embodiment provides a weld seam positioning system for box-type workpieces based on three-dimensional point clouds, including: The weld path reference direction recognition module is configured to acquire the three-dimensional point cloud of the weld area of ​​the target to be welded in the box-type class, calculate the path direction vector based on the centroid of the point cloud of adjacent scanning positions and normalize it to obtain the weld path reference direction. The local plane fitting module is configured to perform plane detection and fitting based on the point cloud data of the target weld area to obtain the fitted plane; and to perform intelligent plane filtering based on the angle threshold between the weld path reference direction and the plane normal vector to obtain multiple candidate planes. The plane selection module is configured to construct a comprehensive evaluation function F, which includes angle score, distance score, flatness score, and thickness score, to filter multiple candidate planes and obtain the target plane; The geometric intersection module is configured to generate a family of planes from the target plane by offsetting the direction of the normal vector, and then perform geometric intersection with the triangular mesh surface of the welding target area to form the weld intersection line; The projection module is configured to project the weld intersection line back to the target plane and resample it to form the final trajectory point sequence used for welding.

[0045] It should be noted that each module in this embodiment corresponds one-to-one with each step in embodiment 1, and their specific implementation process is the same, so it will not be repeated here.

[0046] Example 4 Based on Embodiment 1, this embodiment provides an electronic device, including a memory and a processor, as well as computer instructions stored in the memory and running on the processor. When the computer instructions are executed by the processor, they complete the steps in the method for locating weld seams of box-type workpieces based on three-dimensional point clouds as described in Embodiment 1.

[0047] Example 5 A computer-readable storage medium for storing computer instructions, which, when executed by a processor, complete the steps in the method for locating weld seams of box-type workpieces based on three-dimensional point clouds as described in Embodiment 1.

[0048] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0049] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A method for locating weld seams in box-type workpieces based on 3D point clouds, characterized in that, Includes the following steps: The three-dimensional point cloud of the weld seam area of ​​the target to be welded in the box-type object is obtained. The path direction vector is calculated based on the centroid of the point cloud of adjacent scanning positions and normalized to obtain the reference direction of the weld seam path. Plane detection and fitting are performed based on point cloud data of the target weld area to obtain the fitted plane; intelligent plane filtering is performed based on the angle threshold between the weld path reference direction and the plane normal vector to obtain multiple candidate planes. A comprehensive evaluation function F, including angle score, distance score, flatness score, and thickness score, is constructed to filter multiple candidate planes and obtain the target plane. The target plane is used to generate a family of planes by offsetting the direction of the normal vector, and then geometrically intersected with the triangular mesh surface of the welding target area to form the weld intersection line; The weld intersection line is projected back onto the target plane and resampled to form the final trajectory point sequence used for welding.

2. The method for locating weld seams of box-type workpieces based on three-dimensional point clouds as described in claim 1, characterized in that: The method for obtaining the 3D point cloud of the target weld area, calculating the path direction vector based on the centroids of adjacent scanned point clouds, and normalizing the vector to obtain the weld path reference direction includes the following steps: The target to be welded is scanned from different scanning positions to obtain three-dimensional point cloud data, which is then preprocessed. For point cloud data at different scanning positions, calculate the centroid coordinates of the point cloud at each scanning position, and construct the weld path direction vector from the centroids of the point cloud data obtained from two adjacent frames of scanning and normalize it. The direction corresponding to the largest unit component of the normalized unit path vector is used as the reference direction for the weld path.

3. The method for locating weld seams of box-type workpieces based on three-dimensional point clouds as described in claim 2, characterized in that: Methods for preprocessing the obtained 3D point cloud data include: Perform Z-axis filtering on the point cloud data to remove points that exceed the set range; The filtered point cloud data is downsampled by voxelization, and the point cloud data space is divided into multiple cubic grids, i.e., voxels, with each grid retaining a representative point.

4. The method for locating weld seams of box-type workpieces based on three-dimensional point clouds as described in claim 2, characterized in that: Select the centroid of the point cloud data obtained from two consecutive scan frames, and construct the weld path direction vector using the following formula: ； in, and These are the centroid coordinates of the point cloud data for the previous and next frames, respectively.

5. The method for locating weld seams of box-type workpieces based on three-dimensional point clouds as described in claim 1, characterized in that: Planar detection and fitting are performed based on the point cloud data of the target weld area. An improved RANSAC algorithm is used to perform multi-plane detection and fitting on the point cloud, including the following steps: An initial plane is obtained by fitting an initial plane through random sampling; Based on the consistency of geometric distance and normal, a candidate set of interior points of the initial plane is selected, and interior points are those that belong to the plane. Inward point candidate set Perform connectivity analysis and retain the largest connected component as the final set of interior points. Based on the number of points in the maximum connected component, the set of points in the effective plane is identified, and the least squares method is used for fitting based on the RANSAC algorithm iteratively to obtain the fitted plane.

6. The method for locating weld seams of box-type workpieces based on three-dimensional point clouds as described in claim 1, characterized in that: A method for intelligent plane selection based on an angle threshold between the weld path reference direction and the plane normal vector to obtain multiple candidate planes includes the following steps: Based on a preset plane width threshold, calculate the extreme value difference of the plane point cloud in the X, Y, and Z directions, and filter out planes whose extreme value difference in all three directions is lower than the threshold. Set angle threshold For each weld path direction vector Filter out plane normal vectors With weld path direction vector Planes with an included angle greater than a set angle threshold are selected as candidate planes. Alternatively, when multiple candidate planes exist, a comprehensive evaluation function F is constructed, which includes angle score, distance score, flatness score, and thickness score, as shown in the following formula: in, , , , These are the weighting coefficients, corresponding to the weights of the included angle score, distance score, flatness score, and thickness, respectively. The included angle score reflects the consistency between the plane normal vector and the weld path direction; The distance score reflects the distance from the plane to the outer surface of the box-shaped part; the closer the distance, the higher the score. The flatness score reflects the smoothness of the surface of the plane; the higher the flatness, the higher the score. Score the thickness; Choose the plane with the largest F-value as the final target plane. .

7. A robot, characterized in that: The robot is configured to perform the weld seam positioning method for box-type workpieces based on three-dimensional point clouds as described in any one of claims 1-6, identify the weld seam position, and weld the object to be welded.

8. A weld seam positioning system for box-type workpieces based on three-dimensional point clouds, characterized in that, include: The weld path reference direction recognition module is configured to acquire the three-dimensional point cloud of the weld area of ​​the target to be welded in the box-type class, calculate the path direction vector based on the centroid of the point cloud of adjacent scanning positions and normalize it to obtain the weld path reference direction. The local plane fitting module is configured to perform plane detection and fitting based on the point cloud data of the target weld area to obtain the fitted plane; and to perform intelligent plane filtering based on the angle threshold between the weld path reference direction and the plane normal vector to obtain multiple candidate planes. The plane selection module is configured to construct a comprehensive evaluation function F, which includes angle score, distance score, flatness score, and thickness score, to filter multiple candidate planes and obtain the target plane; The geometric intersection module is configured to generate a family of planes from the target plane by offsetting the direction of the normal vector, and then perform geometric intersection with the triangular mesh surface of the welding target area to form the weld intersection line; The projection module is configured to project the weld intersection line back to the target plane and resample it to form the final trajectory point sequence used for welding.

9. An electronic device, characterized in that, It includes a memory and a processor, as well as computer instructions stored in the memory and running on the processor, which, when executed by the processor, complete the steps in the method for locating weld seams of box-type workpieces based on three-dimensional point clouds as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, Used to store computer instructions, which, when executed by a processor, complete the steps in the method for locating weld seams of box-type workpieces based on three-dimensional point clouds as described in any one of claims 1-7.

Images