Vertical welding seam identification method, computer equipment, readable storage medium and program product

Abstract

The invention relates to a vertical welding seam recognition method, computer equipment, a computer readable storage medium and a computer program product. The method comprises the steps that a target single-workpiece point cloud of a target workpiece is obtained, and the target workpiece comprises a plurality of rib plates and a bottom plate; rib plate point clouds corresponding to the rib plates are extracted from the target single workpiece point clouds; all the rib plate point clouds are projected to the plane where the bottom plate is located, centroid downsampling is conducted on all the projected rib plate point clouds, and multiple welding seam point clouds are obtained; identifying a first linear welding seam point cloud and a second linear welding seam point cloud from the welding seam point clouds, and under the condition that the first linear welding seam point cloud intersects with the second linear welding seam point cloud, detecting the spatial position relationship between the intersection position and the first linear welding seam point cloud and the spatial position relationship between the intersection position and the second linear welding seam point cloud; and generating a welding seam identification result according to the spatial position relationship. By adopting the method, the vertical welding seam identification accuracy can be improved.

Description

Technical Field

[0001] This application relates to the field of welding technology, and in particular to a method for identifying vertical welds, a computer device, a computer-readable storage medium, and a computer program product. Background Technology

[0002] With the continuous evolution of automation and intelligent technologies in shipbuilding, automatic identification and path planning of weld seams in ship assembly structural components have become a core element in improving welding production efficiency and process quality.

[0003] In traditional techniques, weld identification is usually performed based on the acquired 3D point cloud of the workpiece, using planar or linear feature fitting methods. For example, algorithms such as RANSAC (Random Sample Consensus) are used to segment geometric elements representing stiffeners or weld edges from the point cloud, thereby determining the weld location.

[0004] However, when faced with a complex and multi-workpiece assembly scenario, traditional methods often produce a series of isolated two-dimensional planar line segments, which are difficult to effectively construct the three-dimensional spatial topological relationship between welds. In particular, when identifying vertical welds formed by the intersection of stiffening plates and base plates, there are often cases of missed identification, misidentification, or incomplete path information, resulting in a significant lack of accuracy in identifying vertical welds. Summary of the Invention

[0005] Therefore, it is necessary to provide a vertical weld identification method, computer equipment, computer-readable storage medium, and computer program product that can improve the accuracy of vertical weld identification in response to the above-mentioned technical problems.

[0006] In a first aspect, this application provides a method for identifying vertical welds, including:

[0007] Obtain the point cloud of the target single workpiece, wherein the target workpiece includes multiple stiffeners and a base plate;

[0008] Extract the rib point cloud corresponding to each rib from the point cloud of the target single workpiece;

[0009] The point cloud of each stiffener plate is projected onto the plane of the base plate, and the centroid of each point cloud of the projected stiffener plate is downsampled to obtain multiple weld point clouds.

[0010] If a first straight weld point cloud and a second straight weld point cloud are identified from each weld point cloud, and the first straight weld point cloud intersects with the second straight weld point cloud, the spatial positional relationship between the intersection position and the first straight weld point cloud and the second straight weld point cloud is detected.

[0011] Based on spatial relationships, weld identification results are generated.

[0012] Secondly, this application also provides a vertical weld identification device, comprising:

[0013] The acquisition module is used to acquire the point cloud of the target single workpiece, wherein the target workpiece includes multiple stiffeners and a base plate;

[0014] The extraction module is used to extract the rib point cloud corresponding to each rib from the point cloud of the target single workpiece.

[0015] The projection module is used to project the point cloud of each stiffener plate onto the plane of the base plate, and to perform centroid downsampling on the projected point cloud of each stiffener plate to obtain multiple weld point clouds.

[0016] The detection module is used to detect the spatial relationship between the intersection point and the first straight weld point cloud and the second straight weld point cloud when a first straight weld point cloud and a second straight weld point cloud are identified from each weld point cloud and the first straight weld point cloud intersects with the second straight weld point cloud.

[0017] The generation module is used to generate weld recognition results based on spatial location relationships.

[0018] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0019] Obtain the point cloud of the target single workpiece, wherein the target workpiece includes multiple stiffeners and a base plate;

[0020] Extract the rib point cloud corresponding to each rib from the point cloud of the target single workpiece;

[0021] The point cloud of each stiffener plate is projected onto the plane of the base plate, and the centroid of each point cloud of the projected stiffener plate is downsampled to obtain multiple weld point clouds.

[0022] If a first straight weld point cloud and a second straight weld point cloud are identified from each weld point cloud, and the first straight weld point cloud intersects with the second straight weld point cloud, the spatial positional relationship between the intersection position and the first straight weld point cloud and the second straight weld point cloud is detected.

[0023] Based on spatial relationships, weld identification results are generated.

[0024] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:

[0025] Obtain the point cloud of the target single workpiece, wherein the target workpiece includes multiple stiffeners and a base plate;

[0026] Extract the rib point cloud corresponding to each rib from the point cloud of the target single workpiece;

[0027] The point cloud of each stiffener plate is projected onto the plane of the base plate, and the centroid of each point cloud of the projected stiffener plate is downsampled to obtain multiple weld point clouds.

[0028] If a first straight weld point cloud and a second straight weld point cloud are identified from each weld point cloud, and the first straight weld point cloud intersects with the second straight weld point cloud, the spatial positional relationship between the intersection position and the first straight weld point cloud and the second straight weld point cloud is detected.

[0029] Based on spatial relationships, weld identification results are generated.

[0030] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, performs the following steps:

[0031] Obtain the point cloud of the target single workpiece, wherein the target workpiece includes multiple stiffeners and a base plate;

[0032] Extract the rib point cloud corresponding to each rib from the point cloud of the target single workpiece;

[0033] The point cloud of each stiffener plate is projected onto the plane of the base plate, and the centroid of each point cloud of the projected stiffener plate is downsampled to obtain multiple weld point clouds.

[0034] If a first straight weld point cloud and a second straight weld point cloud are identified from each weld point cloud, and the first straight weld point cloud intersects with the second straight weld point cloud, the spatial positional relationship between the intersection position and the first straight weld point cloud and the second straight weld point cloud is detected.

[0035] Based on spatial relationships, weld identification results are generated.

[0036] The aforementioned vertical weld identification method, computer equipment, computer-readable storage medium, and computer program product first acquire a single workpiece point cloud of the target workpiece, which includes multiple stiffeners and a base plate. By extracting the stiffener point clouds corresponding to each stiffener and projecting them onto the plane of the base plate, effective dimensionality reduction and alignment of the complex three-dimensional structure of the target workpiece are achieved. Subsequently, by performing centroid downsampling on the projected stiffener point clouds, key geometric features of the weld can be retained while filtering out noise and redundant data, forming a feature point set characterizing the weld distribution. Further, by identifying the first and second straight-line weld point clouds and determining whether they intersect, accurate determination of the existence of a vertical weld is achieved. If intersecting first and second straight-line weld point clouds exist, a vertical weld is determined to exist. Furthermore, by detecting the spatial relationship between the intersection point and the two straight-line weld point clouds, the projection direction of each stiffener at the intersection and its relative position to the base plate are analyzed, thereby accurately inferring the starting point, direction, and effective extension range of the vertical weld. The vertical weld identification results generated based on this spatial relationship can not only clearly identify the existence of vertical welds, but also provide more complete and continuous spatial path information. This effectively overcomes the technical defects of traditional methods, such as missed or misjudged vertical welds or incomplete path descriptions due to isolated features and missing topological relationships, and significantly improves the accuracy and reliability of vertical weld identification. Attached Figure Description

[0037] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0038] Figure 1 This is a diagram illustrating the application environment of the vertical weld identification method in one embodiment of this application.

[0039] Figure 2 This is a flowchart illustrating an embodiment of the vertical weld identification method of this application;

[0040] Figure 3 This is a flowchart illustrating a vertical weld identification method in another embodiment of this application;

[0041] Figure 4 This is a structural block diagram of an isolated weld identification device in one embodiment of this application;

[0042] Figure 5 This is an internal structural diagram of a computer device in one embodiment of this application. Detailed Implementation

[0043] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0044] Currently, automated welding of ship assembly structures faces challenges across the entire chain, from identification to planning. Traditional operating methods heavily rely on human experience or semi-automatic planning based on two-dimensional drawings, resulting in low efficiency and difficulty in ensuring consistency. Even with the introduction of automation technology, mainstream methods are mostly based on ideal CAD (Computer-Aided Design) models for weld extraction. However, workpieces in actual production often exhibit deformation and assembly deviations, causing a significant decrease in positioning accuracy of these methods in real, complex multi-workpiece scenarios.

[0045] To address the discrepancy between models and physical objects, some research has shifted towards direct analysis based on 3D point clouds. However, existing point cloud processing methods have limitations. They typically focus on simple geometric feature segmentation of individual workpieces (such as fitting planes or lines using the RANSAC algorithm), and the segmentation results are highly unstable when dealing with complex scenarios involving multiple workpieces that are mixed, mutually occluded, and have similar point cloud features. More importantly, these methods mostly only output scattered local line segments or edge fragments, lacking an understanding of the global topology of the weld. This results in fragmented and unconnected outputs, making them unsuitable for directly generating continuous and reliable robotic welding trajectories.

[0046] In multi-workpiece assembly scenarios, existing solutions reveal a series of technical bottlenecks. First, in the initial point cloud processing stage, simple threshold segmentation is easily affected by platform interference, leading to incomplete workpiece point cloud extraction. Second, overlapping projections of multiple workpieces can easily cause mis-clustering, affecting the accurate separation of individual workpieces. Third, in the critical feature extraction stage, traditional filtering and clustering methods lack stability under noise and complex geometry, making it difficult to robustly separate stiffeners from the base plate and segment individual stiffeners themselves. Fourth, endpoint and intersection detection is highly susceptible to noise, leading to missed or false detections. Fifth, the determination of opposing welds often relies on simple geometric thresholds, which cannot handle complex situations such as near-collinearity and short line segments, thus failing to accurately infer the direction, height, and effective range of the opposing weld, ultimately resulting in unreliable automatically generated welding paths.

[0047] The vertical weld identification method provided in this application embodiment can be applied to, for example... Figure 1In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be integrated onto server 104 or placed on the cloud or other network servers. Terminal 102 can be, but is not limited to, welding equipment or its control terminal. The control terminal can be, but is not limited to, various personal computers, laptops, smartphones, tablets, drones, low-altitude aircraft, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, smart vehicle devices, projection equipment, etc. Portable wearable devices can include smartwatches, smart bracelets, head-mounted devices, etc. Head-mounted devices can be virtual reality (VR) devices, augmented reality (AR) devices, smart glasses, etc. Server 104 can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services.

[0048] In one exemplary embodiment, such as Figure 2 As shown, a method for identifying vertical welds is provided, which can be applied to... Figure 1 Taking terminal 102 as an example, the explanation includes the following steps 202 to 210. Wherein:

[0049] Step 202: Obtain the target single workpiece point cloud, wherein the target workpiece includes multiple stiffeners and a base plate.

[0050] In this embodiment, the application scenario is described using the weld seam identification stage before the welding trajectory planning of the small assembled workpiece as an example.

[0051] The target workpiece can refer to the welding object that needs to be identified for weld seam identification.

[0052] In some embodiments, the target tool may refer to a ship assembly component consisting of a base plate and multiple stiffening plates mounted and fixed to the base plate. The base plate may be a steel plate forming the bottom of the hull. The stiffening plates may be plates welded or fixed to the base plate to enhance the rigidity and strength of the structure.

[0053] A weld can refer to a continuous joint on a target workpiece that requires welding operations.

[0054] Welds can include stiffener-base plate welds and vertical welds.

[0055] The stiffening rib-base plate weld refers to the joint between the stiffening rib and the base plate.

[0056] Vertical welds can refer to the joints where two or more stiffening plates meet.

[0057] In some embodiments, the stiffener-base plate weld is located in the plane of the base plate; the vertical weld intersects the plane of the base plate, and one of its endpoints is located in the plane of the base plate.

[0058] A target single workpiece point cloud can refer to a large set of spatial point data that contains only the three-dimensional geometric information of the surface of a single workpiece, obtained after scanning the target workpiece with a three-dimensional vision sensor such as a laser scanner or a structured light 3D camera.

[0059] In some embodiments, in a scenario where multiple small workpieces are placed on a platform and scanned together, the point cloud of a single target workpiece can be obtained by segmentation from the point cloud of multiple workpieces.

[0060] For example, before identifying vertical weld seams, a 3D vision sensor can be used to acquire 3D data of the target workpiece. After preprocessing the acquired raw data, or without preprocessing, a single-workpiece point cloud covering the surface of the target workpiece can be obtained. The single-workpiece point cloud can be directly sent to the terminal for vertical weld seam identification, or it can be stored in a specified storage path first. When the terminal needs to identify vertical weld seams on the target workpiece, it can directly obtain the single-workpiece point cloud from the 3D vision sensor, or it can obtain the single-workpiece point cloud from the specified storage path.

[0061] In some embodiments, the preprocessing of the acquired raw data may include: denoising the raw data by at least one point cloud preprocessing technique such as outlier filtering and statistical filtering, and using segmentation algorithms based on geometric features or deep learning to separate the point cloud portion belonging to the target workpiece from the raw data, removing interfering point clouds from the platform, adjacent workpieces or other backgrounds to obtain the target single workpiece point cloud.

[0062] In some embodiments, the welding robot can be suspended upside down on a ground-rail cantilever, and a 3D vision sensor can be mounted on the ground-rail cantilever, maintaining a constant relative position with the welding robot's base coordinate system. During the actual point cloud acquisition process, the ground-rail cantilever moves at a constant speed along a preset path, synchronously driving the 3D vision sensor to scan the target workpiece.

[0063] In some embodiments, the spatial coordinates of the target single workpiece point cloud can be represented by the base coordinate system of the welding robot. That is, the point cloud data collected by the 3D vision sensor can be accurately transformed into the base coordinate system of the welding robot according to the pre-calibrated coordinate transformation parameters between the 3D vision sensor and the welding robot, so as to facilitate the subsequent generation of weld seam trajectory planning data that can directly drive the robot's movement.

[0064] Step 204: Extract the corresponding rib point cloud from the point cloud of the target single workpiece.

[0065] Among them, the rib point cloud can refer to the subset of point cloud data segmented from the point cloud of the target single workpiece, which corresponds to each rib in the target workpiece.

[0066] For example, a point cloud plane fitting algorithm can be used to fit a planar model representing the base plate from the point cloud of the target single workpiece. Based on the fitted planar model, the distance between each point in the point cloud of the target single workpiece and the base plate plane can be calculated. Points with a distance less than a preset distance threshold are identified as belonging to the base plate, thereby extracting the base plate point cloud; while points with a distance greater than the preset threshold are identified as candidate point clouds for stiffeners. Since the candidate point clouds for stiffeners may contain noise, small fragments, or mixed point clouds of multiple stiffeners, the candidate point clouds for stiffeners can be further segmented to separate the stiffener point clouds corresponding to each stiffener.

[0067] As an example, Euclidean clustering analysis can be performed on the candidate point cloud of stiffeners to divide the point cloud into multiple independent point cloud clusters based on the spatial distance between points. The clustering results are then filtered using prior knowledge (such as the typical size range of stiffeners) to remove excessively small noisy clusters and to identify the point cloud clusters that meet the geometric characteristics as the stiffener point cloud corresponding to an independent stiffener.

[0068] Point cloud plane fitting algorithms can include, but are not limited to, least squares method, principal component analysis and RANSCA (random sample consensus).

[0069] Step 206: Project the point cloud of each stiffener plate onto the plane of the base plate, and perform centroid downsampling on the projected point cloud of each stiffener plate to obtain multiple weld point clouds.

[0070] Among them, weld point cloud can refer to the set of feature points obtained after projection and downsampling processing, which is used to characterize the potential location of the weld.

[0071] Centroid downsampling can refer to calculating the centroid of each rib 2D point cloud cluster formed by a large number of points after projection, and then using the average position of all points in the centroid or its neighborhood (a small search radius can be defined) as the final feature point representing the cluster.

[0072] For example, to eliminate interference in the height direction and focus on the weld layout on the base plate, all points in the point cloud of each stiffener are vertically projected onto the base plate plane. After projection, each stiffener forms a point cloud band or a strip on the plane, and the central axis of this band corresponds to the weld line between the stiffener and the base plate. Then, by downsampling the centroid of each projected point cloud band, the entire point cloud representing the stiffener is condensed into a few key points located on the center line of its projection area; these key points constitute the weld point cloud.

[0073] Step 208: When the first straight weld point cloud and the second straight weld point cloud are identified from each weld point cloud, and the first straight weld point cloud intersects with the second straight weld point cloud, the spatial positional relationship between the intersection position and the first straight weld point cloud and the second straight weld point cloud is detected.

[0074] In this context, a straight weld point cloud can refer to the set of points extracted from each weld point cloud by performing straight-line fitting or identification, representing a continuous straight weld. It is understood that weld point clouds can include not only straight weld point clouds but also curved weld point clouds.

[0075] The first and second straight weld point clouds can be any two intersecting straight weld point clouds from each weld point cloud.

[0076] For example, after obtaining multiple discrete weld point clouds, a line detection algorithm is used to identify multiple straight weld point clouds. Then, each straight weld point cloud is combined in pairs. For any combination of straight weld point clouds, the intersection position of the two straight weld point clouds in the combination is detected. If an intersection position exists, one of the straight weld point clouds in the combination is designated as the first straight weld point cloud, and the other is designated as the second straight weld point cloud. Furthermore, the spatial relationship between the intersection position and the first and second straight weld point clouds is detected.

[0077] Among them, the line detection algorithm may include, but is not limited to, Hough transform, RANSAC line fitting, or point cloud clustering and fitting based on distance threshold.

[0078] The detection of the spatial relationship between the intersection point and the two straight weld point clouds includes, but is not limited to, at least one of the following: determining whether the intersection point falls within the actual line segment interval covered by the two straight weld point clouds; determining whether the intersection point falls at the endpoints of the two straight weld point clouds.

[0079] Step 210: Generate weld identification results based on spatial positional relationships.

[0080] For example, according to the preset vertical weld judgment rules, the existence of a vertical weld is determined based on the detected spatial position relationship, and the judgment result is used as the vertical weld identification result.

[0081] In some embodiments, when a vertical weld is present, the generated vertical weld identification result may also include at least one of the following: the trajectory start point, trajectory direction, weld height, and search vector of the vertical weld.

[0082] In some embodiments, the preset vertical weld determination rule may include: if the intersection is located on the point cloud of two straight welds, then it is determined that there is a vertical weld at that location.

[0083] In the aforementioned method for identifying vertical welds, a single workpiece point cloud containing multiple stiffeners and a base plate is first acquired. By extracting the stiffener point clouds corresponding to each stiffener and projecting them onto the plane of the base plate, effective dimensionality reduction and alignment of the complex three-dimensional structure of the target workpiece are achieved. Subsequently, by performing centroid downsampling on the projected stiffener point clouds, key geometric features of the weld are retained while filtering out noise and redundant data, forming a feature point set representing the weld distribution. Furthermore, by identifying the first and second straight-line weld point clouds and determining whether they intersect, the existence of a vertical weld is accurately determined. If intersecting first and second straight-line weld point clouds exist, a vertical weld is identified. Further analysis of the spatial relationship between the intersection point and the two straight-line weld point clouds, along with the projection direction of each stiffener at the intersection and its relative position to the base plate, allows for accurate inference of the starting point, direction, and effective extension range of the vertical weld. The vertical weld identification results generated based on this spatial relationship can not only clearly identify the existence of vertical welds, but also provide more complete and continuous spatial path information. This effectively overcomes the technical defects of traditional methods, such as missed or misjudged vertical welds or incomplete path descriptions due to isolated features and missing topological relationships, and significantly improves the accuracy and reliability of vertical weld identification.

[0084] In one exemplary embodiment, multiple workpieces are placed together on the platform; such as Figure 3 As shown, obtaining the target single workpiece point cloud includes steps 302 to 310. Wherein:

[0085] Step 302: Obtain the multi-workpiece scan point cloud and the platform template point cloud of the platform, and segment the single-workpiece scan point cloud corresponding to each workpiece from the multi-workpiece scan point cloud.

[0086] It should be noted that when multiple workpieces are randomly placed on the same platform, the point cloud data acquired using 3D scanning simultaneously includes both the workpieces and part of the platform background. Because the workpieces are positioned in different orientations on the platform, the platform itself may have local deformation or tilt, and the top structure of the workpiece (such as tall stiffeners) may cause uneven height distribution, using a preset global height threshold to separate the workpiece from the platform, or filtering based on a single height benchmark after coarse segmentation, is difficult to accurately adapt to the local environment of each workpiece. This can easily lead to incorrect truncation of the workpiece point cloud during filtering, or failure to effectively filter out the platform point cloud below the workpiece, directly affecting the quality of the input data and the final accuracy of subsequent weld seam identification.

[0087] Among them, multi-workpiece scanning point cloud refers to the original three-dimensional point cloud data set obtained by scanning the entire working area containing multiple workpieces and the platform below them at one time using a three-dimensional vision sensor.

[0088] A single workpiece scan point cloud can refer to a subset of point clouds that corresponds one-to-one with each workpiece, obtained by segmenting the scan point clouds of multiple workpieces according to the workpiece using point cloud segmentation technology.

[0089] Platform template point cloud refers to the collection of point cloud data containing only the surface of the platform obtained by performing a 3D scan of an empty platform before placing any workpiece.

[0090] For example, when no workpieces are placed on the platform, a 3D vision sensor mounted on the ground rail cantilever scans the platform to obtain a set of point cloud data containing only the platform surface, i.e., the platform template point cloud. After multiple workpieces to be identified for vertical weld seam recognition are placed on the platform, the 3D vision sensor mounted on the ground rail cantilever scans the platform with multiple workpieces to obtain a multi-workpiece scan point cloud covering the entire multi-workpiece group vertical scene. Subsequently, the point cloud is processed according to a preset point cloud segmentation rule to separate the point cloud regions belonging to different workpieces, and each segmented independent point cloud cluster is taken as a single workpiece scan point cloud.

[0091] As an example, the preset point cloud segmentation rules can employ a spatial distance-based clustering algorithm, such as Euclidean clustering. Based on the three-dimensional Euclidean distance between points in the point cloud, points with a distance less than a set threshold are grouped into the same cluster, thereby separating the spatially independent point clouds of individual workpieces.

[0092] As another example, the preset point cloud segmentation rules can employ a semantic segmentation model based on deep learning. This semantic segmentation model is pre-trained with a large amount of labeled data, enabling it to learn the deep features and contextual information of the workpiece point cloud, achieve end-to-end segmentation of multiple workpiece scan point clouds, directly output the workpiece category or instance to which each point belongs, and thus generate the single workpiece scan point cloud corresponding to each workpiece.

[0093] Step 304: Perform pass-through filtering on the platform template point cloud according to the first pass-through filtering range corresponding to each workpiece to obtain the initial platform point cloud corresponding to each workpiece.

[0094] Pass-through filtering is a spatial filtering method for point clouds. It allows you to retain only points that fall within a specified range along one or more coordinate axes. For example, by setting z_min and z_max in the Z-axis direction, you can filter out points with excessively high or low Z values.

[0095] The maximum value of the first filter sub-range perpendicular to the platform direction in the first direct-pass filter range is the sum of the maximum value of the corresponding single workpiece scan point cloud in the direction perpendicular to the platform and the preset first increment value.

[0096] For example, to facilitate explanation, an XYZ three-dimensional coordinate system can be established, with the direction perpendicular to the platform plane and away from the workpiece as the positive Z-axis. For each single workpiece scan point cloud, its minimum and maximum coordinate values ​​in the X, Y, and Z axes are calculated. To ensure the integrity of the filtered data, a preset first increment value can be added to its maximum coordinate value in the Z-axis direction, and the sum is determined as the maximum value in the Z-axis direction within the first through-filter range corresponding to that workpiece. The minimum value in the Z-axis direction within the first through-filter range corresponding to that workpiece can be the minimum coordinate value of the single workpiece scan point cloud in the Z-axis direction, or it can be unrestricted. The boundary values ​​of the first through-filter range corresponding to that workpiece in the X direction can be the minimum and maximum coordinate values ​​of the single workpiece scan point cloud in the X-axis direction. The boundary values ​​of the first through-filter range corresponding to that workpiece in the Y direction can be the minimum and maximum coordinate values ​​of the single workpiece scan point cloud in the Y-axis direction. Subsequently, each first pass-through filtering range is used as a filtering condition to perform pass-through filtering on the global platform template point cloud. After pass-through filtering, for each workpiece, only the points in the platform template point cloud that fall within its corresponding first pass-through filtering range are retained to form the initial platform segmentation point cloud for each workpiece.

[0097] Step 306: Fit the platform plane corresponding to each workpiece according to each initial platform point cloud, and project each initial platform point cloud onto its corresponding platform plane to obtain the target platform point cloud corresponding to each workpiece.

[0098] For example, for each initial platform point cloud, a plane fitting algorithm is used to fit the platform plane corresponding to each workpiece; then, all points in the initial platform point cloud are vertically projected onto the fitted platform plane to obtain the target platform point cloud corresponding to each workpiece.

[0099] Step 308: Perform pass-through filtering on the single workpiece scanning point cloud corresponding to each workpiece according to the second pass-through filtering range corresponding to each workpiece, and obtain the single workpiece point cloud corresponding to each workpiece.

[0100] The maximum value of the second filter sub-range perpendicular to the platform direction in the second direct filtering range is the sum of the minimum value of the target platform point cloud in the direction perpendicular to the platform and the preset second increment value.

[0101] For example, for each target platform point cloud, its minimum and maximum coordinate values ​​in the X, Y, and Z axes are calculated. To ensure the integrity of the filtered data, a preset second increment value can be added to the minimum coordinate value in the Z-axis direction, and the resulting sum is determined as the maximum value in the Z-axis direction within the second pass-through filtering range corresponding to the workpiece. There is no restriction on the minimum value in the Z-axis direction within the second pass-through filtering range corresponding to the workpiece. The boundary values ​​of the second pass-through filtering range corresponding to the workpiece in the X-axis direction can be the minimum and maximum coordinate values ​​of the single workpiece scan point cloud corresponding to the workpiece in the X-axis direction. The boundary values ​​of the second pass-through filtering range corresponding to the workpiece in the Y-axis direction can be the minimum and maximum coordinate values ​​of the single workpiece scan point cloud corresponding to the workpiece in the Y-axis direction.

[0102] For each workpiece, after determining the second pass-through filtering range corresponding to the workpiece, the second pass-through filtering range is used as the filtering condition to perform a pass-through filtering operation on the single workpiece scan point cloud corresponding to the workpiece, so as to filter out the platform point cloud part in the single workpiece scan point cloud; after pass-through filtering, for each workpiece, only the points in the single workpiece scan point cloud that fall within its corresponding second pass-through filtering range are retained to form the single workpiece point cloud corresponding to each workpiece.

[0103] Step 310: Take any workpiece as the target workpiece, and take the single workpiece point cloud corresponding to the target workpiece as the target single workpiece point cloud.

[0104] For example, from all the workpieces, one is selected as the target workpiece for which weld seam identification is currently required, and its corresponding single workpiece point cloud, after the above processing, is used as the target single workpiece point cloud for subsequent vertical weld seam identification.

[0105] In this embodiment, the relevant platform regions are first extracted from the platform template point cloud using the point cloud range of each workpiece, and a target platform point cloud with higher global accuracy is established through fitting and projection. Then, based on the height benchmark of the target platform point cloud, the boundary between the platform and the workpiece is defined by combining a preset second increment value, effectively separating the workpiece and the platform, and improving the segmentation accuracy and robustness in multi-workpiece scenarios.

[0106] In an exemplary embodiment, segmenting the single-workpiece scan point cloud corresponding to each workpiece from the multi-workpiece scan point cloud includes:

[0107] The multi-workpiece scan point cloud is downsampled to obtain a multi-workpiece downsampled point cloud. Platform background points matching the platform template point cloud are removed from the multi-workpiece downsampled point cloud to obtain the multi-workpiece initial point cloud. The multi-workpiece initial point cloud is projected onto a preset reference horizontal plane, and the projected multi-workpiece initial point cloud is clustered and segmented to obtain the single-workpiece projected point cloud corresponding to each workpiece. The third pass-through filtering range corresponding to each workpiece is determined based on the single-workpiece projected point cloud; the third pass-through filtering range does not constrain the direction perpendicular to the platform. Pass-through filtering is performed on the multi-workpiece scan point cloud according to each first pass-through filtering range to obtain the single-workpiece scan point cloud corresponding to each workpiece.

[0108] It should be noted that directly segmenting a single workpiece from the point cloud of multiple workpieces results in huge computational overhead and is prone to blurring of segmentation boundaries, oversegmentation, or undersegmentation due to uneven point cloud density and interference from platform background points, which seriously affects the accuracy and efficiency of single workpiece point cloud extraction.

[0109] For example, the acquired multi-workpiece scan point cloud can first be downsampled to obtain a multi-workpiece downsampled point cloud. Then, the multi-workpiece downsampled point cloud is registered and aligned with a pre-acquired platform template point cloud. By calculating the distance between the two point clouds and setting a distance threshold, points in the downsampled point cloud whose distance to the platform template point cloud is less than the threshold are identified as platform background points and removed. The remaining point cloud, far from the platform template, can serve as the initial point cloud for the multi-workpiece. Subsequently, all points in the initial point cloud are vertically projected onto a preset reference horizontal plane to obtain its two-dimensional projection point set. Then, distance-based clustering analysis (e.g., Euclidean clustering) is performed on this two-dimensional point set, setting an appropriate distance threshold to cluster projection points that are spatially close to each other and belong to the same workpiece into one class. Each successfully clustered point cloud cluster is the single-workpiece projection point cloud, representing the projection contour of the corresponding workpiece on the platform.

[0110] Then, for each single workpiece's projected point cloud, calculate its minimum and maximum coordinate values ​​in the X, Y, and Z axes. To ensure the integrity of the filtered data, subtract a preset third increment value from the minimum coordinate values ​​in the X and Y axes, and add the preset third increment value to the maximum coordinate values ​​in the X and Y axes to obtain the boundary values ​​in the X and Y axes of the third through-filter range corresponding to each workpiece; no boundary constraints are required in the Z axis direction of the third through-filter range corresponding to each workpiece. For example, for workpiece A1, first calculate the minimum value X_min and maximum value X_max of its single workpiece projected point cloud in the X axis direction; the minimum value Y_min and maximum value Y_max in the Y axis direction; assuming the preset third increment value is 20 mm, the final determined third through-filter range is (X_min-20, X_max+20) in the X axis direction, (Y_min-20, Y_max+20) in the Y axis direction, and no constraints are set in the Z axis direction.

[0111] Then, each third pass-through filter range is used as a filtering condition to perform pass-through filtering on the global multi-workpiece scan point cloud. After pass-through filtering, for each workpiece, only the points in the multi-workpiece scan point cloud that fall within its corresponding third pass-through filter range are retained to form the single-workpiece scan point cloud corresponding to each workpiece.

[0112] In this embodiment, firstly, downsampling of the original point cloud significantly reduces the computational burden of subsequent registration and segmentation. Secondly, by accurately removing the background through comparison with the platform template point cloud, interference from the platform on workpiece segmentation is effectively eliminated. Next, the 3D segmentation problem is transformed into a more manageable 2D planar clustering problem, significantly improving the robustness and accuracy of workpiece instance segmentation. Finally, using the precise planar range derived from the 2D projection results, complete 3D workpiece data is extracted from the original high-resolution point cloud, ensuring the detailed integrity of the final single-workpiece scan point cloud. This method significantly improves processing efficiency while maintaining segmentation accuracy.

[0113] In an exemplary embodiment, before detecting the spatial relationship between the intersection point and the spatial relationship between the first straight weld point cloud and the second straight weld point cloud, when a first straight weld point cloud and a second straight weld point cloud are identified from each weld point cloud, the method further includes:

[0114] When multiple straight weld point clouds are identified from each weld point cloud, the included angle between any two straight weld point clouds is detected; when the included angle between the first straight weld point cloud and the second straight weld point cloud is within the preset intersection angle range, it is determined that the first straight weld point cloud and the second straight weld point cloud intersect, and the first straight weld point cloud is one of the straight weld point clouds, and the second straight weld point cloud is one of the straight weld point clouds.

[0115] It should be noted that the multiple straight weld seam point clouds identified from the discrete weld seam point clouds only represent potential line segments. Directly performing pairwise intersection calculations on all identified straight weld seam point clouds would introduce a large amount of meaningless computation and may lead to misjudgments. For example, two straight weld seam point clouds that are actually parallel or approximately parallel in space and do not intersect at all may have mathematically fictitious intersections at a distance due to fitting errors or local point cloud interference. If such intersections are considered valid without any judgment, it will significantly increase the error rate and computational load of subsequent processing, affecting the overall accuracy and efficiency of vertical weld seam identification.

[0116] In this context, a straight weld point cloud refers to the set of points that represent a continuous straight weld, identified through straight-line fitting or detection of the weld point cloud. As an example, this can be the set of points corresponding to a straight line segment fitted from discrete points using the Hough transform or the RANSAC algorithm.

[0117] The preset intersection angle range can refer to a pre-defined angle threshold range, such as [10°, 170°], used to determine whether two straight lines can intersect within a finite length. If the included angle between two straight weld point clouds falls within this range, they are determined to intersect; if the included angle is close to 0° or 180°, and the two straight weld point clouds are parallel or approximately parallel, they are determined to be impossible to intersect within a finite length. Since the vertical weld is located at the junction of two stiffening plates, and the straight weld point cloud is the projection of the stiffening plate onto the bottom plate plane, for two non-intersecting straight weld point clouds, it can be determined that the two stiffening plates do not intersect, and therefore it can be determined that there is no vertical weld here.

[0118] For example, after successfully identifying multiple straight weld seam point clouds from the weld seam point cloud, these straight line segments are first paired. For each pair of straight weld seam point clouds, the direction vector of the line they represent is calculated, and the angle between these two direction vectors is calculated. The angle can be an acute angle or the smallest angle within the range of 0° to 180°. For each pair of straight weld seam point clouds, the angle is compared with a preset intersection angle range. If the angle between the first and second straight weld seam point clouds is detected to be within the preset intersection angle range, i.e., neither close to 0° nor close to 180°, then it is determined that the two straight weld seams represented by the first and second straight weld seam point clouds intersect. If the angle between the first and second straight weld seam point clouds is detected to be outside the preset intersection angle range, i.e., close to 0° or 180°, then it is determined that the two straight weld seams represented by the first and second straight weld seam point clouds do not intersect.

[0119] In this embodiment, by utilizing the simple angular relationship between two straight lines in a two-dimensional plane, a large number of obviously parallel line combinations that cannot produce effective intersections within a finite length are quickly and efficiently filtered out. This not only significantly reduces unnecessary computational overhead and improves overall processing efficiency, but more importantly, it can preemptively eliminate interference from illusory intersections caused by parallel or near-collinear relationships. This allows subsequent spatial positional relationship analysis to focus on the effective straight line weld point cloud pairs that are truly likely to form vertical welds, thereby improving the accuracy and reliability of the vertical weld identification process.

[0120] In an exemplary embodiment, for each weld point cloud, before detecting the included angle between two straight weld point clouds in the case that multiple straight weld point clouds are identified from each weld point cloud, the method further includes:

[0121] Traverse the points in the weld point cloud and detect the number of endpoints in the weld point cloud; when the number of endpoints in the weld point cloud is zero, determine that the weld point cloud is a closed curve weld point cloud; when the number of endpoints in the weld point cloud is two, construct the endpoint connecting line segment between the two endpoints and detect the interval distance between each weld point and the endpoint connecting line segment; if the average value of each interval distance is less than the preset distance threshold, determine that the weld point cloud is a straight line weld point cloud.

[0122] It should be noted that the method of directly applying Hough transform or RANSAC line fitting for straight weld seam identification has low reliability in cases where there are non-standard point clouds caused by complex geometry or noise (such as slightly curved arc segments, irregularly distributed discrete point sets, etc.), or when multiple point clouds are collinear but sparsely distributed.

[0123] Among them, the endpoint connecting line segment can refer to the straight line segment that connects two endpoints.

[0124] For example, for a weld point cloud, the distribution characteristics of points in its neighborhood can first be analyzed by calculating the K nearest neighbors of each point in the weld point cloud. If all the neighboring points of a point are distributed on one side, while there are no neighboring points on the other side, then the point may be located at an endpoint of the point cloud. By traversing all the points in the weld point cloud and statistically identifying the endpoint positions with this distribution characteristic, the number of endpoints in the weld point cloud can be obtained.

[0125] If the number of endpoints is zero, it indicates that all points in the weld point cloud are in interconnected loops, with no obvious start or end points. In this case, the weld point cloud can be determined to have a closed geometry, and therefore it can be identified as a closed curve weld point cloud.

[0126] If there are exactly two endpoints, the weld point cloud may be either curved or straight. Therefore, the coordinates of these two endpoints can be used to construct an endpoint connecting line segment in space. Then, for each point in the weld point cloud, the vertical distance from it to this endpoint line segment is calculated; this vertical distance is the interval distance between the point and the endpoint connecting line segment. After obtaining the interval distances from all points to the endpoint connecting line segment, the average of these distances is calculated. This average is compared with a preset distance threshold. If the average is less than the preset distance threshold, it indicates that the point cloud is generally closely distributed near the endpoint line segment, conforming to the characteristics of a straight line; therefore, the weld point cloud is determined to be a straight weld point cloud. Conversely, if the average is greater than or equal to the threshold, it indicates that the weld point cloud deviates significantly from a straight line and may be curved or irregularly shaped; therefore, it is determined not to be a straight weld point cloud.

[0127] In this embodiment, the weld point cloud is first preliminarily classified into topologies based on the number of endpoints to quickly eliminate weld point clouds with closed curves. Then, by calculating the average deviation of the line segments connecting all points to the endpoints, the linear and non-linear distributions are quickly and accurately distinguished.

[0128] In one exemplary embodiment, detecting the number of endpoints in the weld point cloud includes:

[0129] Based on the set of directional information from each point in the weld point cloud to other points, at least one candidate endpoint is determined from the weld point cloud; radius clustering is performed on each candidate endpoint to obtain multiple clusters; the geometric center point of each cluster is determined as the endpoint, and the number of endpoints is counted.

[0130] It should be noted that in the weld point cloud obtained from actual scanning, the boundaries of the point cloud may be irregular due to scanning noise, uneven point cloud density, and edge effects of complex geometric structures. Therefore, the neighborhood distribution characteristics of individual points may be unstable. This can easily lead to misjudgments: on the one hand, true endpoints may be missed due to local point cloud sparseness or interference; on the other hand, points located at the boundary but not at the end may be misjudged as endpoints due to similar neighborhood distributions, ultimately resulting in inaccurate endpoint counts and affecting the correct classification of the point cloud as either a straight line or a curve.

[0131] The direction information set can refer to the set of vectors that point from one point to all other points in the weld point cloud. For example, assuming the weld point cloud includes points A, B, and C, the direction information set for point A can be represented as {vector AB, vector AC}.

[0132] Candidate endpoints can refer to points that are considered to have potential endpoint characteristics based on preliminary analysis of directional information.

[0133] A cluster can refer to a group of points that are spatially aggregated after being grouped using a radius clustering algorithm.

[0134] The geometric center point can be the point determined by the average of the coordinates of all points in a point cloud cluster, representing the spatial center of the cluster.

[0135] For example, for each point in the weld point cloud, the direction vector pointing to all other points in the weld point cloud is calculated to form a set of direction information corresponding to that point. Analyzing this set of direction information, if the spatial angle range covered by the direction vector of that point exceeds a preset angle threshold, such as exceeding 180°, then the point may be located inside the point cloud; if the spatial angle range covered by the direction vector of that point does not exceed the preset angle threshold, such as not exceeding 180°, then the point may be located at the end of the point cloud, and the point can be identified as a candidate endpoint.

[0136] Due to noise or irregular point cloud boundaries, multiple candidate endpoints with similar directional features may exist near the endpoint region. Therefore, these candidate endpoints can be further clustered by radius. If the distance between two or more candidate endpoints is less than the preset clustering radius, they can be grouped into the same cluster. For each cluster, the average coordinates of all points within the cluster are calculated to obtain the geometric center point of the cluster. The geometric center point of each cluster is then used as an endpoint, and the number of endpoints is counted.

[0137] In this embodiment, the directional information set is first used for preliminary screening of endpoints, reducing false alarms caused by relying solely on local neighborhoods. Subsequently, spatially adjacent candidate endpoints are merged through radius clustering, and the geometric center of the cluster is used as the final endpoint. This not only improves the accuracy of endpoint location but also removes duplicate candidate points, reducing interference from multiple false endpoints caused by irregular point cloud boundaries or noisy points.

[0138] In an exemplary embodiment, generating a weld identification result based on spatial location relationships further includes:

[0139] When the intersection point coincides with any endpoint of the first straight weld seam point cloud and also coincides with any endpoint of the second straight weld seam point cloud, a first search vector is constructed with the intersection point as the starting point and the other endpoint of the first straight weld seam point cloud that does not coincide with the intersection point as the ending point. A second search vector is constructed with the intersection point as the starting point and the other endpoint of the second straight weld seam point cloud that does not coincide with the intersection point as the ending point. A first height is determined based on the first straight weld seam point cloud and the first search vector, and a second height is determined based on the second straight weld seam point cloud and the second search vector. If both the first height and the second height are greater than a preset height threshold, the number of vertical welds is determined to be one. The intersection point is taken as the trajectory starting point of the vertical weld, and the smaller of the first height and the second height is taken as the height of the vertical weld. A weld recognition result is generated based on the trajectory starting point, the height of the vertical weld, the first search vector, and the second search vector.

[0140] For example, if, based on spatial relationships, the intersection point coincides with any endpoint of the first straight weld seam point cloud and also coincides with any endpoint of the second straight weld seam point cloud, a first search vector is constructed starting from the intersection point and ending at the other endpoint of the first straight weld seam point cloud that does not coincide with the intersection point. Similarly, a second search vector is constructed starting from the intersection point and ending at the other endpoint of the second straight weld seam point cloud that does not coincide with the intersection point. Then, for each upper surface point in the first straight weld seam point cloud, its vertical distance to the first search vector is calculated. The calculated vertical distances are sorted in ascending order, and the average of the first preset number of vertical distances is calculated, which is then determined as the first height. Similarly, for each upper surface point in the second straight weld seam point cloud, its vertical distance to the second search vector is calculated. The calculated vertical distances are sorted in ascending order, and the average of the first preset number of vertical distances is calculated, which is then determined as the second height.

[0141] Subsequently, the first height and the second height are compared with preset height thresholds. If both the first height and the second height are greater than the preset height thresholds, it is determined that a vertical weld exists at this location, and the number of vertical welds is one. The intersection point is taken as the trajectory starting point of the vertical weld, and the smaller of the first height and the second height is taken as the height of the vertical weld. The trajectory starting point, the height of the vertical weld, the first search vector, and the second search vector are used as the content of the vertical weld identification result to generate the vertical weld identification result.

[0142] Among them, the upper surface point refers to the point on the surface farthest from the base plate.

[0143] In some embodiments, the vertical weld identification result may also include the vertical weld direction.

[0144] In an exemplary embodiment, generating a weld identification result based on spatial location relationships further includes:

[0145] If the intersection point coincides with any endpoint of the first straight weld seam point cloud, and the intersection point does not coincide with any endpoint of the second straight weld seam point cloud, a third search vector is constructed starting from the intersection point and ending at the other endpoint of the first straight weld seam point cloud that does not coincide with the intersection point; a fourth search vector is constructed starting from the intersection point and ending at any endpoint of the second straight weld seam point cloud that does not coincide with the intersection point; a fifth search vector is constructed starting from the intersection point and ending at the other endpoint of the second straight weld seam point cloud; a third height is determined based on the first straight weld seam point cloud and the third search vector; a fourth height is determined based on the second straight weld seam point cloud and the fourth search vector; and a fifth height is determined based on the second straight weld seam point cloud. The fifth height is determined by the fifth search vector; if the third, fourth, and fifth heights are all greater than the preset height threshold, the number of vertical welds is determined to be two; the intersection position is taken as the trajectory starting point of the vertical weld, the smaller of the third and fourth heights is taken as the height of the first vertical weld, and the smaller of the third and fifth heights is taken as the height of the second vertical weld; the first vertical weld identification sub-result is generated based on the trajectory starting point, the height of the first vertical weld, the third search vector, and the fourth search vector, and the second vertical weld identification sub-result is generated based on the trajectory starting point, the height of the second vertical weld, the fourth search vector, and the fifth search vector, and the first and second vertical weld identification sub-results are taken as the vertical weld identification result.

[0146] For example, if, based on spatial relationships, it is determined that the intersection point coincides with any endpoint of the first straight weld point cloud, and the intersection point does not coincide with any endpoint of the second straight weld point cloud, a third search vector is first constructed with the intersection point as the starting point and the other endpoint of the first straight weld point cloud that does not coincide with the intersection point as the ending point; a fourth search vector is constructed with the intersection point as the starting point and any endpoint of the second straight weld point cloud that does not coincide with the intersection point as the ending point; and a fifth search vector is constructed with the intersection point as the starting point and the other endpoint of the second straight weld point cloud that does not coincide with the intersection point as the ending point. Then, for each upper surface point in the first straight weld seam point cloud, calculate its vertical distance to the third search vector, sort the calculated vertical distances in ascending order, calculate the average of the first preset number of vertical distances in the sorted order, and determine the average as the third height; for each upper surface point in the second straight weld seam point cloud, calculate its vertical distance to the fourth search vector, sort the calculated vertical distances in ascending order, calculate the average of the first preset number of vertical distances in the sorted order, and determine the average as the fourth height; for each upper surface point in the second straight weld seam point cloud, calculate its vertical distance to the fifth search vector, sort the calculated vertical distances in ascending order, calculate the average of the first preset number of vertical distances in the sorted order, and determine the average as the fifth height.

[0147] Subsequently, the third, fourth, and fifth heights are compared with preset height thresholds. If all three heights are greater than the preset height thresholds, it is determined that a vertical weld exists at this location, and the number of vertical welds is two. The intersection point is taken as the trajectory starting point of the vertical weld. The smaller of the third and fourth heights is taken as the height of the first vertical weld, and the smaller of the third and fifth heights is taken as the height of the second vertical weld. The trajectory starting point, the height of the first vertical weld, the third search vector, and the fourth search vector are used as the content of the first vertical weld identification sub-result to generate the first vertical weld identification sub-result. The trajectory starting point, the height of the second vertical weld, the fourth search vector, and the fifth search vector are used as the content of the second vertical weld identification sub-result to generate the second vertical weld identification sub-result. The first and second vertical weld identification sub-results are then merged to obtain the vertical weld identification result.

[0148] In an exemplary embodiment, generating a weld identification result based on spatial location relationships further includes:

[0149] When the intersection point does not coincide with any endpoint of the first straight weld seam point cloud, and the intersection point does not coincide with any endpoint of the second straight weld seam point cloud, a sixth search vector is constructed starting from the intersection point and ending at any endpoint of the first straight weld seam point cloud that does not coincide with the intersection point; a seventh search vector is constructed ending at the other endpoint of the first straight weld seam point cloud that does not coincide with the intersection point. Similarly, an eighth search vector is constructed starting from the intersection point and ending at any endpoint of the second straight weld seam point cloud that does not coincide with the intersection point; a ninth search vector is constructed ending at the other endpoint of the second straight weld seam point cloud that does not coincide with the intersection point. A sixth height is determined based on the first straight weld seam point cloud and the sixth search vector; a seventh height is determined based on the first straight weld seam point cloud and the seventh search vector; an eighth height is determined based on the second straight weld seam point cloud and the eighth search vector; and a ninth height is determined based on the second straight weld seam point cloud and the ninth search vector. If the sixth, seventh, eighth, and ninth heights are all greater than a preset height threshold, a vertical weld seam is determined. The number is four; the intersection position is taken as the trajectory starting point of the vertical weld, the smaller of the sixth and seventh heights is taken as the height of the third vertical weld, the smaller of the seventh and eighth heights is taken as the height of the fourth vertical weld, the smaller of the eighth and ninth heights is taken as the height of the fifth vertical weld, and the smaller of the ninth and sixth heights is taken as the height of the sixth vertical weld; the third vertical weld identification sub-result is generated based on the trajectory starting point, the height of the third vertical weld, the sixth search vector, and the seventh search vector; the fourth vertical weld identification sub-result is generated based on the trajectory starting point, the height of the fourth vertical weld, the seventh search vector, and the eighth search vector; the fifth vertical weld identification sub-result is generated based on the trajectory starting point, the height of the fifth vertical weld, the eighth search vector, and the ninth search vector; the sixth vertical weld identification sub-result is generated based on the trajectory starting point, the height of the fifth vertical weld, the ninth search vector, and the sixth search vector; the third, fourth, fifth, and sixth vertical weld identification sub-results are taken as the vertical weld identification results.

[0150] For example, if, based on spatial relationships, it is determined that the endpoints of the intersecting positions do not coincide with the endpoints of the first straight weld point cloud, and the endpoints of the intersecting positions do not coincide with the endpoints of the second straight weld point cloud, a sixth search vector is constructed, starting from the intersecting position and ending at any endpoint in the first straight weld point cloud that does not coincide with the intersecting position; a seventh search vector is constructed, starting from the intersecting position and ending at the other endpoint in the first straight weld point cloud that does not coincide with the intersecting position; an eighth search vector is constructed, starting from the intersecting position and ending at any endpoint in the second straight weld point cloud that does not coincide with the intersecting position; and a ninth search vector is constructed, starting from the intersecting position and ending at the other endpoint in the second straight weld point cloud that does not coincide with the intersecting position. Then, for each upper surface point in the first straight weld seam point cloud, calculate its vertical distance to the sixth search vector, sort the calculated vertical distances in ascending order, calculate the average of the first preset number of vertical distances in the sorted order, and determine the average of this average as the sixth height; for each upper surface point in the first straight weld seam point cloud, calculate its vertical distance to the seventh search vector, sort the calculated vertical distances in ascending order, calculate the average of the first preset number of vertical distances in the sorted order, and determine the average of this average as the seventh height; for each upper surface point in the second straight weld seam point cloud, calculate its vertical distance to the eighth search vector, sort the calculated vertical distances in ascending order, calculate the average of the first preset number of vertical distances in the sorted order, and determine the average of this average as the eighth height; for each upper surface point in the second straight weld seam point cloud, calculate its vertical distance to the ninth search vector, sort the calculated vertical distances in ascending order, calculate the average of the first preset number of vertical distances in the sorted order, and determine the average of this average as the ninth height.

[0151] Subsequently, the sixth, seventh, eighth, and ninth heights are compared with preset height thresholds. If all of these heights are greater than the preset height threshold, it is determined that a vertical weld exists at this location, and the number of vertical welds is four. The intersection point is taken as the trajectory starting point of the vertical weld. The smaller of the sixth and seventh heights is taken as the height of the third vertical weld, the smaller of the seventh and eighth heights as the height of the fourth vertical weld, the smaller of the eighth and ninth heights as the height of the fifth vertical weld, and the smaller of the ninth and sixth heights as the height of the sixth vertical weld. The trajectory starting point, the height of the third vertical weld, the sixth search vector, and the seventh search vector are used as the content of the third vertical weld identification sub-result to generate the third vertical weld identification sub-result. The trajectory starting point, the height of the fourth vertical weld, the seventh search vector, and the eighth search vector are used as the content of the third vertical weld identification sub-result. The vector is used as the content of the fourth vertical weld identification sub-result to generate the fourth vertical weld identification sub-result; the trajectory starting point, the height of the fifth vertical weld, the eighth search vector, and the ninth search vector are used as the content of the fifth vertical weld identification sub-result to generate the fifth vertical weld identification sub-result; the trajectory starting point, the height of the sixth vertical weld, the ninth search vector, and the sixth search vector are used as the content of the sixth vertical weld identification sub-result to generate the sixth vertical weld identification sub-result; the third, fourth, fifth, and sixth vertical weld identification sub-results are fused to obtain the vertical weld identification result.

[0152] In this embodiment, the number of potential vertical welds is automatically and accurately inferred based on the typical positional relationship between the intersection point and the endpoint of the straight line. Simultaneously, by constructing a search vector starting from the intersection point and logically combining it, all potential welding directions and planar information that conform to geometric constraints and originate from that intersection point are automatically enumerated and generated. This improves the completeness and physical rationality of the welding path, enabling the robot to effectively plan the scanning and welding trajectory based on this result, significantly enhancing the automation level of vertical weld recognition and the reliability of path planning.

[0153] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0154] Based on the same inventive concept, this application also provides a vertical weld seam identification device for implementing the vertical weld seam identification method described above. The solution provided by this device is similar to the implementation described in the above method; therefore, the specific limitations of one or more vertical weld seam identification device embodiments provided below can be found in the limitations of the vertical weld seam identification method described above, and will not be repeated here.

[0155] In one exemplary embodiment, such as Figure 4 As shown, a vertical weld seam identification device is provided, including: an acquisition module 402, an extraction module 404, a projection module 406, a detection module 408, and a generation module 410, wherein:

[0156] The acquisition module 402 is used to acquire the target single workpiece point cloud of the target workpiece, wherein the target workpiece includes multiple stiffeners and a base plate;

[0157] Extraction module 404 is used to extract the rib point cloud corresponding to each rib from the point cloud of the target single workpiece.

[0158] Projection module 406 is used to project the point cloud of each stiffener plate onto the plane of the base plate, and to perform centroid downsampling on the projected point cloud of each stiffener plate to obtain multiple weld point clouds.

[0159] The detection module 408 is used to detect the spatial relationship between the intersection point and the first straight weld point cloud and the second straight weld point cloud when the first straight weld point cloud and the second straight weld point cloud are identified from each weld point cloud and the first straight weld point cloud intersect with the second straight weld point cloud.

[0160] The generation module 410 is used to generate weld recognition results based on spatial positional relationships.

[0161] In one exemplary embodiment, multiple workpieces are placed together on the platform; the acquisition module 402 is further configured to:

[0162] The process involves acquiring multi-workpiece scan point clouds and a platform template point cloud, and segmenting individual workpiece scan point clouds from the multi-workpiece scan point clouds. Based on the first direct-pass filtering range corresponding to each workpiece, a direct-pass filtering is applied to the platform template point cloud to obtain the initial platform point cloud for each workpiece. The initial platform point clouds are then fitted to the corresponding platform planes of each workpiece, and projected onto their respective platform planes to obtain the target platform point cloud for each workpiece. A second direct-pass filtering range is then applied to the individual workpiece scan point cloud for each workpiece to obtain the individual workpiece point cloud for each workpiece. Any workpiece is designated as the target workpiece, and the individual workpiece point cloud corresponding to the target workpiece is designated as the target individual workpiece point cloud. The maximum value of the first filtering sub-range perpendicular to the platform direction within the first direct-pass filtering range is the sum of the maximum value of the corresponding individual workpiece scan point cloud in the direction perpendicular to the platform and a preset first increment value. The maximum value of the second filtering sub-range perpendicular to the platform direction within the second direct-pass filtering range is the sum of the minimum value of the target platform point cloud in the direction perpendicular to the platform and a preset second increment value.

[0163] In an exemplary embodiment, the acquisition module 402 is further configured to:

[0164] The multi-workpiece scan point cloud is downsampled to obtain a multi-workpiece downsampled point cloud. Platform background points matching the platform template point cloud are removed from the multi-workpiece downsampled point cloud to obtain the multi-workpiece initial point cloud. The multi-workpiece initial point cloud is projected onto a preset reference horizontal plane, and the projected multi-workpiece initial point cloud is clustered and segmented to obtain the single-workpiece projected point cloud corresponding to each workpiece. The third pass-through filtering range corresponding to each workpiece is determined based on the single-workpiece projected point cloud; the third pass-through filtering range does not constrain the direction perpendicular to the platform. Pass-through filtering is applied to the multi-workpiece scan point cloud according to each third pass-through filtering range to obtain the single-workpiece scan point cloud corresponding to each workpiece.

[0165] In an exemplary embodiment, the vertical weld seam identification device further includes an identification module; when a first straight weld seam point cloud and a second straight weld seam point cloud are identified from each weld seam point cloud, and the first straight weld seam point cloud intersects with the second straight weld seam point cloud, before detecting the spatial positional relationship between the intersection point and the first straight weld seam point cloud and the second straight weld seam point cloud, the identification module is configured to:

[0166] When multiple straight weld point clouds are identified from each weld point cloud, the included angle between any two straight weld point clouds is detected; when the included angle between the first straight weld point cloud and the second straight weld point cloud is within the preset intersection angle range, it is determined that the first straight weld point cloud and the second straight weld point cloud intersect, and the first straight weld point cloud is one of the straight weld point clouds, and the second straight weld point cloud is one of the straight weld point clouds.

[0167] In an exemplary embodiment, for each weld point cloud, before detecting the included angle between two straight weld point clouds in the case that multiple straight weld point clouds are identified from each weld point cloud, the identification module is further configured to:

[0168] Traverse the points in the weld point cloud and detect the number of endpoints in the weld point cloud; when the number of endpoints in the weld point cloud is zero, determine that the weld point cloud is a closed curve weld point cloud; when the number of endpoints in the weld point cloud is two, construct the endpoint connecting line segment between the two endpoints and detect the interval distance between each weld point and the endpoint connecting line segment; if the average value of each interval distance is less than the preset distance threshold, determine that the weld point cloud is a straight line weld point cloud.

[0169] In one exemplary embodiment, the identification module is further configured to:

[0170] Based on the set of directional information from each point in the weld point cloud to other points, at least one candidate endpoint is determined from the weld point cloud; radius clustering is performed on each candidate endpoint to obtain multiple clusters; the geometric center point of each cluster is determined as the endpoint, and the number of endpoints is counted.

[0171] In one exemplary embodiment, the generation module 410 is further configured to:

[0172] When the intersection point coincides with any endpoint of the first straight weld seam point cloud and also coincides with any endpoint of the second straight weld seam point cloud, a first search vector is constructed with the intersection point as the starting point and the other endpoint of the first straight weld seam point cloud that does not coincide with the intersection point as the ending point; a second search vector is constructed with the intersection point as the starting point and the other endpoint of the second straight weld seam point cloud that does not coincide with the intersection point as the ending point; a first height is determined based on the first straight weld seam point cloud and the first search vector, and a second height is determined based on the second straight weld seam point cloud and the second search vector; if both the first height and the second height are greater than a preset height threshold, the number of vertical welds is determined to be one; the intersection point is taken as the trajectory starting point of the vertical weld, and the smaller of the first height and the second height is taken as the height of the vertical weld; a weld recognition result is generated based on the trajectory starting point, the height of the vertical weld, the first search vector, and the second search vector.

[0173] In one exemplary embodiment, the generation module 410 is further configured to:

[0174] When the intersection point coincides with any endpoint of the first straight weld seam point cloud, and the intersection point does not coincide with any endpoint of the second straight weld seam point cloud, a third search vector is constructed starting from the intersection point and ending at the other endpoint of the first straight weld seam point cloud that does not coincide with the intersection point. A fourth search vector is constructed starting from the intersection point and ending at any endpoint of the second straight weld seam point cloud that does not coincide with the intersection point. A fifth search vector is constructed starting from the intersection point and ending at the other endpoint of the second straight weld seam point cloud. A third height is determined based on the first straight weld seam point cloud and the third search vector. A fourth height is determined based on the second straight weld seam point cloud and the fourth search vector. The fifth search vector determines the fifth height; if the third, fourth, and fifth heights are all greater than the preset height threshold, the number of vertical welds is determined to be two; the intersection position is taken as the trajectory starting point of the vertical weld, the smaller of the third and fourth heights is taken as the height of the first vertical weld, and the smaller of the third and fifth heights is taken as the height of the second vertical weld; the first vertical weld identification sub-result is generated based on the trajectory starting point, the height of the first vertical weld, the third search vector, and the fourth search vector, and the second vertical weld identification sub-result is generated based on the trajectory starting point, the height of the second vertical weld, the fourth search vector, and the fifth search vector, and the first and second vertical weld identification sub-results are taken as the vertical weld identification result.

[0175] In one exemplary embodiment, the generation module 410 is further configured to:

[0176] When the intersection point does not coincide with any endpoint of the first straight weld seam point cloud, and the intersection point does not coincide with any endpoint of the second straight weld seam point cloud, a sixth search vector is constructed starting from the intersection point and ending at any endpoint of the first straight weld seam point cloud that does not coincide with the intersection point; a seventh search vector is constructed starting from the intersection point and ending at the other endpoint of the first straight weld seam point cloud that does not coincide with the intersection point; an eighth search vector is constructed starting from the intersection point and ending at any endpoint of the second straight weld seam point cloud that does not coincide with the intersection point; a ninth search vector is constructed starting from the intersection point and ending at the other endpoint of the second straight weld seam point cloud that does not coincide with the intersection point; a sixth height is determined based on the first straight weld seam point cloud and the sixth search vector; a seventh height is determined based on the first straight weld seam point cloud and the seventh search vector; an eighth height is determined based on the second straight weld seam point cloud and the eighth search vector; and a ninth height is determined based on the second straight weld seam point cloud and the ninth search vector; if the sixth, seventh, eighth, and ninth heights are all greater than a preset height threshold... The number of vertical welds is determined to be four. The intersection position is taken as the trajectory starting point of the vertical weld. The smaller of the sixth and seventh heights is taken as the height of the third vertical weld, the smaller of the seventh and eighth heights is taken as the height of the fourth vertical weld, the smaller of the eighth and ninth heights is taken as the height of the fifth vertical weld, and the smaller of the ninth and sixth heights is taken as the height of the sixth vertical weld. Based on the trajectory starting point, the height of the third vertical weld, the sixth search vector, and the seventh search vector, a sub-result for identifying the third vertical weld is generated. Based on the trajectory starting point, the height of the fourth vertical weld, the seventh search vector, and the eighth search vector, a sub-result for identifying the fourth vertical weld is generated. Based on the trajectory starting point, the height of the fifth vertical weld, the eighth search vector, and the ninth search vector, a sub-result for identifying the fifth vertical weld is generated. Based on the trajectory starting point, the height of the fifth vertical weld, the ninth search vector, and the sixth search vector, a sub-result for identifying the sixth vertical weld is generated. The sub-results for identifying the third, fourth, fifth, and sixth vertical welds are taken as the vertical weld identification results.

[0177] Each module in the aforementioned vertical weld seam identification device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0178] In one exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 5As shown, the computer device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, Near Field Communication (NFC), or other technologies. When executed by the processor, the computer program implements a vertical weld seam identification method. The display unit is used to form a visually visible image and can be a display screen, projection device, or virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0179] Those skilled in the art will understand that Figure 5 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0180] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0181] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.

[0182] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0183] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0184] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, database, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0185] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0186] The above embodiments are merely illustrative of several implementation methods of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for identifying vertical welds, characterized in that, The method includes: Obtain the target single workpiece point cloud of the target workpiece, wherein the target workpiece includes multiple stiffeners and a base plate; Extract the rib point cloud corresponding to each rib from the point cloud of the target single workpiece; Each stiffener point cloud is projected onto the plane of the base plate, and the centroid of each projected stiffener point cloud is downsampled to obtain multiple weld point clouds. If a first straight weld point cloud and a second straight weld point cloud are identified from each of the weld point clouds, and the first straight weld point cloud intersects with the second straight weld point cloud, the spatial positional relationship between the intersection position and the first straight weld point cloud and the second straight weld point cloud is detected. Based on the spatial relationship, a weld identification result is generated.

2. The method according to claim 1, characterized in that, Multiple workpieces are placed together on the platform; The acquisition of the target single workpiece point cloud includes: The system acquires multiple workpiece scan point clouds and the platform template point cloud of the platform, and segments out the single workpiece scan point cloud corresponding to each workpiece from the multiple workpiece scan point clouds. Based on the first pass-through filtering range corresponding to each workpiece, the platform template point cloud is subjected to pass-through filtering to obtain the initial platform point cloud corresponding to each workpiece. Based on the initial platform point cloud of each workpiece, fit the platform plane corresponding to each workpiece, and project the initial platform point cloud of each workpiece onto the corresponding platform plane to obtain the target platform point cloud corresponding to each workpiece. According to the second pass-through filtering range corresponding to each workpiece, the single workpiece scanning point cloud corresponding to each workpiece is passed through filtering to obtain the single workpiece point cloud corresponding to each workpiece. Take any of the aforementioned workpieces as the target workpiece, and take the single workpiece point cloud corresponding to the target workpiece as the target single workpiece point cloud; Wherein, the maximum value of the first filter sub-range perpendicular to the platform direction in the first pass-through filter range is: the sum of the maximum value of the corresponding single workpiece scan point cloud in the direction perpendicular to the platform and a preset first increment value; the maximum value of the second filter sub-range perpendicular to the platform direction in the second pass-through filter range is: the sum of the minimum value of the target platform point cloud in the direction perpendicular to the platform and a preset second increment value.

3. The method according to claim 2, characterized in that, The step of segmenting the single-workpiece scan point cloud corresponding to each workpiece from the multi-workpiece scan point cloud includes... The multi-workpiece scan point cloud is downsampled to obtain a multi-workpiece downsampled point cloud; The platform background points that match the platform template point cloud are removed from the multi-workpiece downsampled point cloud to obtain the multi-workpiece initial point cloud; The initial point cloud of the multiple workpieces is projected onto a preset reference horizontal plane, and the projected initial point cloud of the multiple workpieces is clustered and segmented to obtain the single workpiece projection point cloud corresponding to each workpiece. The third pass-through filtering range corresponding to each workpiece is determined based on the projected point cloud of each workpiece, and the third pass-through filtering range is not constrained in the direction perpendicular to the platform; According to the third pass-through filtering range, the multi-workpiece scanning point cloud is subjected to pass-through filtering to obtain the single-workpiece scanning point cloud corresponding to each of the workpieces.

4. The method according to claim 1, characterized in that, Before detecting the spatial relationship between the intersection point and the spatial position between the first straight weld point cloud and the second straight weld point cloud, in the case where a first straight weld point cloud and a second straight weld point cloud are identified from the weld point clouds, the method further includes: In the case where multiple straight weld point clouds are identified from each of the weld point clouds, the included angle between any two of the straight weld point clouds is detected. When the angle between the first straight weld point cloud and the second straight weld point cloud is within a preset intersection angle range, it is determined that the first straight weld point cloud and the second straight weld point cloud intersect. The first straight weld point cloud is one of the aforementioned straight weld point clouds, and the second straight weld point cloud is one of the aforementioned straight weld point clouds.

5. The method according to claim 4, characterized in that, For each weld point cloud, before detecting the included angle between two of the weld point clouds, in the case that multiple straight weld point clouds are identified from each weld point cloud, the method further includes: Traverse the points in the weld point cloud and detect the number of endpoints in the weld point cloud; When the number of endpoints in the weld point cloud is zero, the weld point cloud is determined to be a closed curve weld point cloud; When the number of endpoints in the weld point cloud is two, an endpoint connection line segment is constructed between the two endpoints, and the interval distance between each weld point and the endpoint connection line segment is detected. If the average value of each of the aforementioned interval distances is less than a preset distance threshold, the weld point cloud is determined to be a straight weld point cloud.

6. The method according to claim 5, characterized in that, The detection of the number of endpoints in the weld point cloud includes: Based on the set of directional information from each point in the weld point cloud to other points, at least one candidate endpoint is determined from the weld point cloud; Radius clustering is performed on each of the candidate endpoints to obtain multiple clusters; The geometric center point of each cluster is determined as the endpoint, and the number of the endpoints is counted.

7. The method according to any one of claims 1 to 6, characterized in that, The step of generating weld identification results based on the spatial positional relationship includes: If the intersection point coincides with any endpoint of the first straight weld seam point cloud and also coincides with any endpoint of the second straight weld seam point cloud, a first search vector is constructed with the intersection point as the starting point and the other endpoint of the first straight weld seam point cloud that does not coincide with the intersection point as the ending point; a second search vector is constructed with the intersection point as the starting point and the other endpoint of the second straight weld seam point cloud that does not coincide with the intersection point as the ending point. The first height is determined based on the first straight weld point cloud and the first search vector, and the second height is determined based on the second straight weld point cloud and the second search vector; If both the first height and the second height are greater than the preset height threshold, the number of vertical welds is determined to be one. The intersection position is taken as the starting point of the vertical weld trajectory, and the smaller of the first height and the second height is taken as the height of the vertical weld. A weld identification result is generated based on the trajectory starting point, the height of the vertical weld, the first search vector, and the second search vector.

8. The method according to any one of claims 1 to 6, characterized in that, The step of generating weld identification results based on the spatial positional relationship includes: If the intersection point coincides with any endpoint of the first straight weld seam point cloud, and the intersection point does not coincide with any endpoint of the second straight weld seam point cloud, a third search vector is constructed with the intersection point as the starting point and the other endpoint of the first straight weld seam point cloud that does not coincide with the intersection point as the ending point; a fourth search vector is constructed with the intersection point as the starting point and any endpoint of the second straight weld seam point cloud that does not coincide with the intersection point as the ending point; and a fifth search vector is constructed with the intersection point as the starting point and the other endpoint of the second straight weld seam point cloud that does not coincide with the intersection point as the ending point. The third height is determined based on the first straight weld point cloud and the third search vector; the fourth height is determined based on the second straight weld point cloud and the fourth search vector; and the fifth height is determined based on the second straight weld point cloud and the fifth search vector. If the third height, the fourth height, and the fifth height are all greater than the preset height threshold, the number of vertical welds is determined to be two. The intersection position is taken as the starting point of the vertical weld trajectory, the smaller of the third height and the fourth height is taken as the height of the first vertical weld, and the smaller of the third height and the fifth height is taken as the height of the second vertical weld. A first vertical weld identification sub-result is generated based on the trajectory starting point, the height of the first vertical weld, the third search vector, and the fourth search vector. A second vertical weld identification sub-result is generated based on the trajectory starting point, the height of the second vertical weld, the fourth search vector, and the fifth search vector. The first vertical weld identification sub-result and the second vertical weld identification sub-result are used as the vertical weld identification result.

9. The method according to any one of claims 1 to 6, characterized in that, The step of generating weld identification results based on the spatial positional relationship includes: If the intersection point does not coincide with any endpoint of the first straight weld seam point cloud, and the intersection point does not coincide with any endpoint of the second straight weld seam point cloud, a sixth search vector is constructed starting from the intersection point and ending at any endpoint of the first straight weld seam point cloud that does not coincide with the intersection point; a seventh search vector is constructed starting from the intersection point and ending at the other endpoint of the first straight weld seam point cloud that does not coincide with the intersection point; an eighth search vector is constructed starting from the intersection point and ending at any endpoint of the second straight weld seam point cloud that does not coincide with the intersection point; and a ninth search vector is constructed starting from the intersection point and ending at the other endpoint of the second straight weld seam point cloud that does not coincide with the intersection point. The sixth height is determined based on the first straight weld point cloud and the sixth search vector; the seventh height is determined based on the first straight weld point cloud and the seventh search vector; the eighth height is determined based on the second straight weld point cloud and the eighth search vector; and the ninth height is determined based on the second straight weld point cloud and the ninth search vector. If the sixth height, the seventh height, the eighth height, and the ninth height are all greater than the preset height threshold, the number of vertical welds is determined to be four. The intersection position is taken as the starting point of the vertical weld trajectory. The smaller of the sixth height and the seventh height is taken as the height of the third vertical weld. The smaller of the seventh height and the eighth height is taken as the height of the fourth vertical weld. The smaller of the eighth height and the ninth height is taken as the height of the fifth vertical weld. The smaller of the ninth height and the sixth height is taken as the height of the sixth vertical weld. A third weld seam identification sub-result is generated based on the trajectory starting point, the height of the third weld seam, the sixth search vector, and the seventh search vector. A fourth weld seam identification sub-result is generated based on the trajectory starting point, the height of the fourth weld seam, the seventh search vector, and the eighth search vector. A fifth weld seam identification sub-result is generated based on the trajectory starting point, the height of the fifth weld seam, the eighth search vector, and the ninth search vector. A sixth weld seam identification sub-result is generated based on the trajectory starting point, the height of the fifth weld seam, the ninth search vector, and the sixth search vector. The third, fourth, fifth, and sixth weld seam identification sub-results are used as the weld seam identification results.

10. A vertical weld seam identification device, characterized in that, The device includes: The acquisition module is used to acquire the point cloud of the target single workpiece, wherein the target workpiece includes multiple stiffeners and a base plate; The extraction module is used to extract the rib point cloud corresponding to each rib from the point cloud of the target single workpiece. The projection module is used to project the point cloud of each stiffener plate onto the plane of the base plate, and to perform centroid downsampling on the projected point cloud of each stiffener plate to obtain multiple weld point clouds. The detection module is used to detect the spatial relationship between the intersection point and the first straight weld point cloud and the second straight weld point cloud when a first straight weld point cloud and a second straight weld point cloud are identified from each weld point cloud and the first straight weld point cloud intersects with the second straight weld point cloud. The generation module is used to generate weld recognition results based on spatial location relationships.

11. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 9.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 9.

13. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 9.

Images