WO2023103621A1 - 一种船舶焊缝特征识别方法 - Google Patents

一种船舶焊缝特征识别方法 Download PDF

Info

Publication number
WO2023103621A1
WO2023103621A1 PCT/CN2022/127298 CN2022127298W WO2023103621A1 WO 2023103621 A1 WO2023103621 A1 WO 2023103621A1 CN 2022127298 W CN2022127298 W CN 2022127298W WO 2023103621 A1 WO2023103621 A1 WO 2023103621A1
Authority
WO
WIPO (PCT)
Prior art keywords
joint
triangle
contour
distance
contour line
Prior art date
Application number
PCT/CN2022/127298
Other languages
English (en)
French (fr)
Inventor
袁明新
戴现令
刘超
孙宏伟
刘锁东
高云强
张本顺
王占光
申燚
Original Assignee
江苏科技大学
中国船舶集团有限公司第七一六研究所
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 江苏科技大学, 中国船舶集团有限公司第七一六研究所 filed Critical 江苏科技大学
Publication of WO2023103621A1 publication Critical patent/WO2023103621A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning

Definitions

  • the invention belongs to the technical field of pattern recognition, and relates to a weld feature recognition technology based on a digital model of a weldment, in particular to a ship weld feature recognition method.
  • Welding is an important link in hull manufacturing. Due to the existence of welding fume, arc light and metal spatter, the welding working environment is often relatively harsh. With the advent of Made in China 2025, the use of welding robots to carry out automated welding can not only improve the consistency of welding quality, but also improve production efficiency and reduce costs, which is the key to promoting the intelligent manufacturing of Chinese ships.
  • weld feature recognition has always been a research focus in the field of automatic welding.
  • the research mainly focuses on the processing of weld image or digital point cloud model, which mainly collects weld image or point cloud model through visual sensor, and then obtains weld feature points based on extraction algorithm, and then realizes the recognition of groove.
  • the secondary development of the weld model is carried out by using the API interface of the modeling software, and the attribute characteristics of the weldment can be obtained, and the weld characteristics can also be identified.
  • weld feature recognition is mainly aimed at a single type of weld, or for the extraction of some joints or individual grooves, which limits its application range.
  • weld seam characteristics are also an important reference for the welding process database to recommend process parameters. Therefore, carrying out the research on ship weld feature recognition will help to realize the accurate recognition of weld features in the digital model of ship weldments, improve the speed and accuracy of welding robot welding process selection, and also increase the number of welding process databases. Versatility and availability for optimizing weld seam formation quality.
  • the purpose of the present invention is to provide a ship weld feature recognition method for the weld feature recognition of the digital model of the weldment.
  • a ship weld feature recognition method firstly generate the STL file of the three-dimensional model through the Delaunay triangulation based on the direction of the direction of the triangle; then carry out the weld feature recognition based on the STL file, and then identify the joint form and groove type.
  • the specific steps are as follows:
  • Step 1 Delaunay triangulation based on concave vertices and triangle handing
  • Step 2 generate the STL file of the three-dimensional model
  • Step 3 extracting the contour lines and points of the model
  • Step 4 Joint space position judgment
  • Delaunay triangulation based on concave vertices and triangle handedness described in step 1 includes initial triangular mesh generation and triangular mesh local transformation, wherein the specific content and steps of the initial triangular mesh generation are as follows:
  • step (6) Determine whether the number of vertices in the array space is less than or equal to 3? If so, end the triangular mesh division, otherwise obtain the newly formed oriented triangle in step (4) or step (5), and then go to step (3) to judge the direction of rotation.
  • the method for the local transformation of the triangular mesh is to remove the boundary edge of the initial triangle division, and process each remaining non-boundary edge according to the following steps:
  • step 2 1) Traversing the initial triangular mesh, finding two adjacent triangles with non-boundary sides as common sides, and judging the concavity and convexity of the quadrilateral formed by these two triangles. When it is a concave quadrilateral, do not do any processing; otherwise, go to step 2);
  • the specific content and method of generating the STL file of the 3D model in step 2 is to obtain the geometric information of the triangular surface by traversing the information of each contour surface of the 3D model and using the above-mentioned Delaunay triangulation meshing algorithm.
  • the STL file in ASCII code file format can be obtained by triangulating the entire model.
  • the specific content and method of extracting the contour lines and points of the model described in step 3 are to combine all triangular faces in the STL file in pairs, and process each pair of triangular faces combined as follows:
  • step (I) Judging whether there are two common vertices in a pair of triangular faces? If so, they are adjacent; otherwise, they are not adjacent. For adjacent triangular faces, then calculate the angle between the normal vectors of the two, and go to step (II); if not adjacent, then take the next pair of triangular faces and continue the adjacent judgment of this step;
  • step (III) Outputting the common line segments of all triangular patches located in different planes in step (II), the entire outline of the model can be extracted;
  • step 4 the judgment of the spatial position of the joint described in step 4 is mainly based on the number of times that the normal vector at the judgment surface is parallel, vertical, and opposite in direction, and the following steps are specifically adopted:
  • step (b) Perform a pairwise cross product on the normal vector obtained in step (a) and obtain the modulus length of the vector product. If the modulus length is 0, then the two normal vectors are parallel. Statistical normal vector appears parallel times P a ;
  • step (c) Perform pairwise dot products on the normal vectors obtained in step (a) and obtain the quantity product. If the product is 0, the two normal vectors are perpendicular. Statistical normal vector appears vertical times V t ;
  • step (d) Carry out two-two direction judgments to the normal vectors obtained in step (a), and count the number of times O p in opposite directions;
  • weld feature identification in step 5 mainly includes joint form identification and groove type identification, wherein the method for joint form identification specifically adopts the following steps:
  • step (B) Take the contour line with the second shortest length as the distance threshold.
  • the shortest contour line group output in step (A) calculate all the distances between the endpoints of any pair of contour lines, and compare the minimum distance with the distance threshold after removing the length of the contour line itself. If the minimum distance is greater than the distance threshold, the contour distance is called a large distance; otherwise, it is called a small distance;
  • the method for groove type identification specifically adopts the following steps:
  • step (D) Identify the groove type according to the total number of contour lines and the joint form. If the total number of contour lines is 24, when the joint form is butt joint or corner joint, go to step (E); when it is lap joint or T-joint, the joints are not bevelled. If the total number of contour lines is 27, go to step (F). If the total number of contour lines is 30, go to step (G);
  • the Delaunay triangulation based on concave vertices and triangle rotation proposed by the present invention can realize fast triangulation of complex polygons such as "comb", and the number of distortion units is small, and the triangular mesh shape is better;
  • the concavo-convexity of each vertex can be obtained effectively by the angle coefficient method, which effectively reduces the time consumed by the initial triangular mesh division and improves the execution efficiency of the algorithm; finally, after the local transformation based on the minimum angle maximum criterion, the mesh is effectively reduced.
  • the degree of grid deformity makes the overall grid closer to a regular triangle, which ensures the generation of better grid quality and is conducive to the acquisition of STL files.
  • the weld seam feature recognition method proposed by the present invention not only recognizes more joint forms and groove types, but also has shorter average running time for feature recognition, and has the advantages of high efficiency and high accuracy.
  • the algorithm only needs to obtain the STL file of the model weld, which has a small amount of pre-processing compared with other methods, which greatly improves the practicability and universality of the algorithm.
  • it also provides a guarantee for the ship welding robot to quickly and accurately select the welding process parameters and improve the welding quality and efficiency.
  • Fig. 1 is a general flow chart of a ship weld feature recognition method of the present invention
  • FIG. 3 Schematic diagram of concave-convex vertex judgment
  • Fig. 4 is a schematic diagram of determining the direction of rotation of a triangle, Figure (a) indicates counterclockwise rotation, and Figure (b) indicates clockwise rotation;
  • Fig. 5 is a schematic diagram for judging a concave vertex inside a triangle
  • Figure (a) indicates that the concave vertex is inside the triangle
  • Figure (b) indicates that the concave vertex is on the edge of the triangle
  • Figure (c) indicates that the concave vertex is outside the triangle
  • Fig. 6 is a schematic diagram of determining the concavo-convexity of a quadrilateral, Figure (a) represents a convex quadrilateral, and Figure (b) represents a concave quadrilateral;
  • Fig. 7 Schematic diagram of minimum angle maximization criterion
  • Figure 8 shows a schematic diagram of a judgment surface
  • Figure (a) shows a schematic diagram of a flat joint
  • Figure (b) shows a schematic diagram of a corner joint
  • Figure (c) shows a schematic diagram of a T-shaped joint
  • figure (a) represents the first 8 groups of contour line schematic diagrams
  • figure (b) represents the first 10 groups of contour line schematic diagrams
  • figure (c) represents the first 12 groups of contour line schematic diagrams
  • Fig. (a) represents the test result of the singly-connected polygon ring tracking algorithm
  • Fig. (b) represents the test result of the triangulation algorithm based on concave-convex vertices
  • Fig. (c) represents the test result of the dynamic Delaunay triangulation algorithm Result
  • figure (d) represents the test result of the present invention
  • Figure 12 The weld feature recognition test of the cabin section, Figure 12(a) shows the case number of the weld in the cabin section, Figure 12(b), Figure 12(c), Figure 12(d), Figure 12(e) Figure 12(f ), Fig. 12(g) represent the characteristics of welds 1, 2, 3, 4, 5 and 6, respectively.
  • a weld feature recognition method of the present invention determines the concave-convex vertices of the polygon by the angle coefficient method, based on the Delaunay triangulation of the concave vertices and the direction of the triangle, construct the triangular mesh on the surface of the 3D model and generate the STL file; then based on the angle between the normal vectors of the adjacent triangular faces, The outline and points of the model are extracted; finally, the weld features are identified according to the spatial position of the joint and the minimum outline distance. Specifically take the following steps:
  • the Delaunay triangulation based on concave vertices and triangle handedness includes initial triangular mesh generation and triangular mesh local transformation, wherein the initial triangular mesh
  • the method of grid generation adopts the following steps:
  • intersection points of P i+1 and P i+2 are A and B.
  • the slopes of the straight lines AP i+1 and BP i+1 be tan ⁇ and tan ⁇ respectively, and calculate the angle coefficient difference ⁇ . If: (i) intersection points A and B are on the same horizontal straight line and ⁇ >0, then the vertex P i+1 is a concave vertex; otherwise, ⁇ 0, then the vertex P i+1 is a convex vertex.
  • intersection points A and B are not on the same horizontal line and ⁇ >0, then the vertex P i+1 is a convex vertex; otherwise, if ⁇ 0, then the vertex P i+1 is a concave vertex. It is calculated as follows:
  • (x 0 , y 0 ) is the coordinate of vertex P j ;
  • (x 1 , y 1 ) is the coordinate of vertex P j+1 ;
  • (x 2 , y 3 ) is the coordinate of vertex P j+2 .
  • step (6) Determine whether the number of vertices in the array space is less than or equal to 3? , if yes, then end the triangular mesh division; otherwise, obtain the directed triangle newly formed in step (4) or step (5), that is, ⁇ P i P i+2 P i+3 or ⁇ P i+1 P i+ 2 P i+3 , and turn to step (3) to judge the direction of rotation.
  • the local transformation of the triangular mesh is based on the minimum angle maximum criterion, so that the triangle shape of the mesh division is optimal, that is, the boundary edge of the initial triangle is removed, and the remaining Each non-boundary edge is processed in the following steps:
  • the common side of two adjacent triangles is recorded as a straight line l 1
  • the straight line connecting two vertices not on the common side of the two triangles is recorded as l 2 . If the straight line l 1 intersects l 2 , it is a convex quadrilateral; otherwise, it is a concave quadrilateral;
  • the method for generating the STL file of the three-dimensional model is mainly to traverse each contour surface of the three-dimensional model by calling the SolidWorks software API function, and extract the normal vector (pointing to the outside of the entity) and the contour vertex coordinate information on each contour surface.
  • the STL file in ASCII code file format can be obtained by triangulating the entire model.
  • the positional relationship between the surfaces to be determined of the two joints directly affects the form of the joint. Since the area of the surface to be determined is relatively large relative to the end surface of the groove, calculate the area of each contour surface and take the largest first four groups to obtain the surface to be determined; then according to the parallel or vertical The spatial position relationship of the four types of joints was discriminated.
  • contour lines of the 10 kinds of welding grooves can usually be divided into three categories: 24, 27 and 30.
  • 24 contour lines there are mainly I-shaped grooves and V-shaped grooves for butt joints; no grooves and single-sided grooves for corner joints; non-grooved grooves for T-shaped and lap joints types.
  • Figure 9(a) is the I-type bevel corresponding to the butt joint in the 24 contour line types.
  • the welding seam feature recognition involved in this type is mainly to take out the first 8 groups of contour lines with the shortest length; then calculate all the distances between the endpoints of any two contour lines for the 8 groups of contour lines; finally, according to the joint spatial position and the minimum contour line distance , to identify the joint form and groove type.
  • Figure 9(b) is the bilateral groove corresponding to the fillet joint among the 27 contour line types.
  • the identification of weld seam features involved in this type is mainly aimed at the shortest 10 sets of contour lines, and referring to the same processing method in the 24 contour lines, to identify the corresponding joint form and groove type.
  • Figure 9(c) is the bilateral groove corresponding to the T-joint among the 30 contour line types.
  • Weld seam feature recognition is mainly aimed at 12 groups of shortest contour lines and referring to the above-mentioned same method to identify the joint form and groove type.
  • the method for extracting the outline and points of the model is to combine all the triangular faces in the STL file in pairs, and to process each pair of triangular faces that are combined as follows:
  • the judgment of the spatial position of the joint is mainly based on the number of times that the normal vector at the judgment surface is parallel, vertical and opposite in direction, and the following steps are specifically adopted:
  • (l, m, n) is the normal vector of the contour surface Coordinates; (o,p,q) is the normal vector of the contour surface coordinate.
  • step (3) Perform pairwise dot products on the normal vectors obtained in step (1) and obtain the quantity product. If the product is 0, the two normal vectors are perpendicular. Counts the number of times V t that the normal vector appears vertical. It is calculated as follows:
  • (l, m, n) is the normal vector of the contour surface Coordinates; (o,p,q) is the normal vector of the contour surface coordinate.
  • step (1) Perform pairwise judgments on the normal vectors obtained in step (1), and count the number of times Opposite directions O p .
  • the weld feature identification mainly includes joint form identification and groove type identification
  • the joint form identification mainly includes the identification of four types of joint forms: butt joint, lap joint, fillet joint and T-joint, and the following steps are adopted:
  • step (2) Take the contour line with the second shortest length as the distance threshold.
  • the shortest contour line group output in step (1) calculate all the distances between the endpoints of any two contour lines, and compare the minimum distance with the distance threshold after removing the length of the contour line itself. If the minimum distance is greater than the distance threshold, the contour distance is called a large distance; otherwise, it is called a small distance;
  • the groove type identification mainly includes I-type, V-type and Y-type grooves of butt joints; non-opening, unilateral and double-sided grooves of corner joints and T-shaped joints, and non-opening grooves of lap joints10 To identify a groove type, the following steps are taken:
  • step (2) Identify the groove type according to the total number of contour lines and the joint form. If the total number of contour lines is 24, when the joint form is a butt joint or a corner joint, go to step (2); when it is a lap joint or a T-joint, the joints are not bevelled. If the total number of contour lines is 27, go to step (3). If the total number of contour lines is 30, go to step (4);
  • Groove identification is performed according to the joint form and the contour line distance. If it is a butt joint with a small spacing, it is a Y-shaped groove for a butt joint; if it is a T-joint with a large spacing, it is a double-sided groove for a T-joint.
  • polygons such as "comb” were subdivided and tested, and combined with the single-connected polygon ring tracking algorithm (single-connected tracking method) [1] , based on The algorithm for triangulation of concave-convex vertices (Bump Vertices) [2] and the dynamic Delaunay triangulation algorithm (Dynamic Delaunay) [3] were compared.
  • the subdivision results of two types of polygons by the four methods are shown in Fig. 10 and Fig. 11.
  • the performance index of subdivision comparison adopts grid average quality coefficient q M and grid correlation quality coefficient q J .
  • the number of deformed triangles is the largest in the single-connected tracking method, and the mesh shape is poor, while the algorithm of the present invention has a small number of deformed triangles compared with it, which is obviously optimized.
  • the dynamic Delaunay method cannot realize the division, but the algorithm of the present invention has a normal division test and is better than the other two methods in the control of the number of deformed triangles, and the division quality is the best.
  • Table 1 shows the grid quality analysis results. It can be seen from the table that whether it is the average quality coefficient of the grid or the associated quality coefficient of the grid, the subdivision quality of the algorithm of the present invention is better than that of the other three methods, and for the polygons in Figure 10, respectively An average increase of 12.49% and 12.78%, and an average increase of 11.62% and 11.74% for the polygons in Figure 11 respectively, thus further illustrating that the algorithm of the present invention effectively reduces the degree of grid deformity, making the entire grid closer to a regular triangle, ensuring Better mesh quality results.
  • Table 2 shows the test results of the three welding seam feature recognition methods. It can be seen from the table that the traversal feature method can only identify two types of butt joints and T-joints, and cannot identify the groove type, and the identification effect is poor .
  • the feature library design method is the same as the algorithm of the present invention, it can accurately identify 6 weld features, but the average running time of joint and groove recognition is higher than the algorithm of the present invention, and the execution efficiency is low.
  • the feature library design The method needs to carry out feature modeling design and establish a database for all welds on the cabin section first, which greatly increases the development workload, and when the database scale is large, it may take a long time to match the weld features.
  • the algorithm of the present invention has a shorter average running time for joint and groove recognition, higher execution efficiency, and only needs to obtain the STL file of the model weld, with a small amount of processing in the early stage, and many joint forms and groove types to be recognized at the same time , high accuracy.
  • Table 3 and Table 4 are the detailed test results of joints and grooves of the algorithm of the present invention respectively. It can be seen from the table that the algorithm of the present invention can quickly count the number of parallel, vertical, and opposite normal vectors at the surface to be determined, and accurately obtain the minimum contour Line distance, distance threshold, total number of contour lines and the angle between the two minimum distance contour lines, and make correct judgments and comparisons based on key parameters, and accurately identify the joint form and groove type of 6 welds, that is, butt joint I-shaped grooves for butt joints, V-shaped grooves for butt joints, no grooves for T-shaped joints, no grooves for corner joints, double-sided grooves for corner joints, and double-sided grooves for T-shaped joints, so as to realize the optimization of weld characteristics Accurate identification and high execution efficiency verify the effectiveness of the algorithm of the present invention.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Artificial Intelligence (AREA)
  • Health & Medical Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Computing Systems (AREA)
  • Databases & Information Systems (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Medical Informatics (AREA)
  • Software Systems (AREA)
  • Image Analysis (AREA)

Abstract

一种船舶焊缝特征识别方法。首先通过角系数法判断多边形的凹凸顶点，基于凹顶点和三角形旋向的Delaunay三角剖分，构造三维模型表面的三角形网格并生成STL文件；然后基于相邻三角面片的法向量夹角，提取出模型的轮廓线及点；接着根据带判定面处法向量进行接头空间位置关系判断；最后根据接头空间位置和最小轮廓线距离完成接头形式和坡口类型的焊缝特征识别。该方法有助于实现船舶焊接件数字模型中焊缝特征的精确识别，进而提高焊接机器人焊接工艺选择的快速性和准确性，达到优化焊接质量、提高生产效率的目的。

Description

一种船舶焊缝特征识别方法 技术领域
本发明属于模式识别技术领域，涉及基于焊接件数字模型的焊缝特征识别技术，尤其涉及一种船舶焊缝特征识别方法。
背景技术
焊接是船体制造中的重要环节，由于焊接烟尘、弧光和金属飞溅等的存在，焊接工作环境往往比较恶劣。随着中国制造2025的到来，利用焊接机器人开展自动化焊接，不仅能提升焊接质量的一致性，还可提高生产效率、降低成本，是推进中国船舶智能化制造的关键。
焊接质量的好坏，其焊接工艺参数的快速、准确选择是关键，而这需要先对焊接件数字模型中焊缝特征进行精确识别。焊缝特征识别一直是自动化焊接领域的研究重点。目前研究主要集中在对焊缝图像或数字点云模型的处理，主要是通过视觉传感器采集到焊缝图像或者点云模型，然后基于提取算法获得焊缝特征点，进而实现坡口的识别。此外，利用建模软件API接口进行焊缝模型的二次开发，获取焊接件的属性特征，也可识别出焊缝特征。但目前焊缝特征识别主要针对单一类型的焊缝，或是针对部分接头或个别坡口的提取，限制了其应用范围。此外，焊缝特征也是焊接工艺数据库进行工艺参数推荐的重要参考依据。因此，开展船舶焊缝特征识别的研究，有助于实现船舶焊接件数字模型中焊缝特征的精确识别，提高焊接机器人焊接工艺选择的快速性和准确性，同时，也增加了焊接工艺数据库的多样性和有效性，有利于优化焊缝成形质量。
发明内容
本发明的目的是针对焊接件数字模型的焊缝特征识别，提供一种船舶焊缝特征识别方法。
为了达到上述目的，本发明采取如下技术方案予以实现：
一种船舶焊缝特征识别方法，首先通过基于有向三角形旋向的Delaunay三角剖分来生成三维模型的STL文件；然后基于STL文件进行焊缝特征识别，进而识别出接头形式和坡口类型。其具体步骤如下：
步骤1.基于凹顶点和三角形旋向的Delaunay三角剖分；
步骤2.生成三维模型的STL文件；
步骤3.提取模型的轮廓线及点；
步骤4.接头空间位置判断；
步骤5.焊缝特征识别。
进一步，步骤1所述基于凹顶点和三角形旋向的Delaunay三角剖分，包括初始三角网格生成和三角网格局部变换，其中，所述初始三角网格生成的具体内容和步骤如下：
(1)逆时针顺序读入多边形顶点，并依次存储到数组空间；
(2)通过角系数法判断每个顶点的凹凸性，获取所有凹顶点；
(3)按顺序从数组空间中依次取出三个顶点来构成有向三角形，求出其面积用于三角形旋向的判断。若计算结果为正，则该三角形为逆时针旋向并转步骤(4)；反之为顺时针旋向并转步骤(5)。若三个顶点位于同一直线上，则也转步骤(5)；
(4)判断三角形内部是否含有凹顶点。若无凹顶点，则输出构成该三角形的三顶点P i、P i+1、P i+2，并将P i+1从数组空间中移除，同时以该三角形的起点P i、终点P i+2和下一个顶点P i+3，构成新有向三角形△P iP i+2P i+3并转步骤(6)；若有凹顶点，则转步骤(5)；
(5)将三角形ΔP iP i+2P i+3的起点P i存回数组空间，取出下一个顶点P i+1，重新构成新有向三角形△P i+1P i+2P i+3并转步骤(6)；
(6)判断数组空间内顶点个数小于等于3？若是，则结束三角形网格划分，否则获取步骤(4)或步骤(5)中新构成的有向三角形，再转步骤(3)进行旋向判断。
所述三角网格局部变换的方法是，去除掉初始划分三角形的边界边，对剩余的每一条非边界边按以下步骤处理：
1)遍历初始三角形网格，找出以非边界边为公共边的两个相邻三角形，并判断由这两个三角形所构成四边形的凹凸性。当为凹四边形，不作任何处理；否则转步骤2)；
2)计算两三角形中的六个内角并取出最小角α，再进行对角线交换，并计算两新三角形中六个内角的最小角β；
3)比较α、β两角大小。当α≥β，保留对角线交换前两三角形；反之保留对角线交换后两新三角形。
进一步，步骤2所述生成三维模型的STL文件的具体内容和方法是，通过遍历三维模型的每个轮廓面信息，利用上述Delaunay三角网格剖分算法，获取三角面片的几何信息。通过对整个模型三角剖分即可得到ASCII码文件格式的STL文件。
进一步，步骤3所述提取模型的轮廓线及点的具体内容和方法是，对STL文件中所有的三角面片两两组合，对组合成的每一对三角面片按如下步骤进行处理：
(Ⅰ)判断一对三角面片是否存在两个公共顶点？若是，则相邻；反之不相邻。对于相邻三角面片，则计算出两者的法向量夹角，并转步骤(Ⅱ)；若不相邻，则取下一对三角面片继续本步骤的相邻判断；
(Ⅱ)根据法向量夹角确定两三角面片的公共线段是否为模型轮廓线。若θ＝0或θ＝180°，则两个公共顶点所组成的公共线段不是模型轮廓线；若0<θ<180°，则公共线段是模型轮廓线的一部分；
(Ⅲ)输出步骤(Ⅱ)中所有位于不同平面内的三角面片公共线段，即可提取到模型的全部轮廓；
(Ⅳ)统计各三角面片顶点在全部轮廓线条中出现的次数n。若n≥3，则该点为模型端点，进而获得所有轮廓顶点。
进一步，步骤4所述接头空间位置判断，主要是基于带判定面处法向量平行、垂直和方向相反的次数进行判断，具体采用如下步骤：
(a)基于海伦公式计算STL中三角面片面积，并求出每个轮廓面面积。取面积最大的前四组，并获取对应轮廓面的法向量；
(b)对步骤(a)获取的法向量进行两两叉积并求出向量积的模长。若模长为0，则两个法向量平行。统计法向量出现平行的次数P a；
(c)对步骤(a)获取的法向量进行两两点积并求出数量积。若积为0，则两个法向量垂直。统计法向量出现垂直的次数V t；
(d)对步骤(a)中获取的法向量进行两两方向判断，并统计出方向相反的次数O p；
(e)根据P a,V t,O p的值判断接头空间位置关系。若P a＝6且O p＝4，则两接头处于平行状态；若V t＝4且O p＝2，则两接头处于垂直状态。
进一步，步骤5所述焊缝特征识别，主要包括接头形式识别和坡口类型识别，其中，所述接头形式识别的方法，具体采用如下步骤：
(A)根据轮廓线总条数获取各类最短轮廓线组。若为24条，则输出前8组长度最短的轮廓线组；若为27条，则输出前10组最短轮廓线组；若为30条，则输出前12组最短轮廓线组；
(B)取次短长度的轮廓线作为距离阈值。针对步骤(A)中输出的最短轮廓线组，计算两两轮廓线的端点间所有距离，去除轮廓线本身长度后，将最小距离与距离阈值对比。若最小距离大于距离阈值，则称轮廓线距离为大间距；反之称为小间距；
(C)根据接头空间位置和轮廓线距离进行接头形式识别。若接头处于平行状态，当轮廓线距离为小间距时，则为平接接头；反之为搭接接头。若接头处于垂直状态，当轮廓线距离为小间距时，则为角接接头；反之为T型接头。
所述坡口类型识别的方法，具体采用如下步骤：
(D)根据轮廓线总条数和接头形式进行坡口类型识别。若轮廓线总条数为24条，当接头形式为平接或角接时，转步骤(E)；当为搭接或T型接头时，则接头都为不开坡口。若轮廓线总条数为27条，转步骤(F)。若轮廓线总条数为30条，转步骤(G)；
(E)根据两条最小距离轮廓线的夹角α进行坡口识别。若接头形式为平接，当α＝0时，则为I型坡口；反之为V型坡口。若接头形式为角接，当α＝90°时，则为不开坡口；反之为单边坡口；
(F)根据轮廓线距离进行坡口识别。若为小间距，则为角接接头的双边坡口；反之为T型接头的单边坡口；
(G)根据接头形式和轮廓线距离进行坡口识别。若为平接接头且小间距，则为平接接头的Y型坡口；若为T型接头且大间距，则为T型接头的双边坡口。
本发明具有以下优点和有益效果：
(1)本发明所提出的基于凹顶点和三角形旋向的Delaunay三角剖分，能实现对如“梳子”等复杂多边形的快速三角剖分，且畸变单元数少，三角网格形态较优；通过角系数法能有效地获取每个顶点的凹凸性，有效减少了初始三角网格划分所消耗的时间，提高了算法的执行效率；最后基于最小角最大准则的局部变换后，有效降低了网格畸形程度，使得全体网格更接近于正三角形，保证了较好网格质量的产生，有利于STL文件的获取。
(2)本发明所提出的模型轮廓线及点的提取，直接通过相邻三角面片的法向量夹角，来判断其公共线段是否为模型轮廓线，有效降低了算法的计算复杂度，实现模型轮廓线及点的快速获取。同时，基于海伦公式的接头空间位置关系判断，实现了焊接接头的初步识别，再结合轮廓线总条数和最小轮廓线距离，即可快速准确识别出4类接头形式和10种坡口类型，进一步缩短了算法运行时间，提高了特征识别算法的准确率和运行效率。
(3)本发明所提出的一种焊缝特征识别方法，不仅识别的接头形式和坡口类型多， 而且在特征识别的平均运行时间上更短，具有高效、高准确率的优势。此外，算法只需要获取模型焊缝的STL文件，相比其他方法前期处理量较小，大大提高了算法的实用性、广泛性。同时，也为船舶焊接机器人快速、准确选择焊接工艺参数，提高焊接质量和效率提供了保障。
附图说明
图1本发明的一种船舶焊缝特征识别方法的总流程图；
图2初始三角网格生成的流程图；
图3凹凸顶点判断的原理图；
图4三角形旋向判定示意图，图(a)表示逆时针旋向，图(b)表示顺时针旋向；
图5三角形内部含有凹顶点判定示意图，图(a)表示凹顶点在三角形内部，图(b)表示凹顶点在三角形边上，图(c)表示凹顶点在三角形外部；
图6四边形凹凸性判定示意图，图(a)表示凸四边形，图(b)表示凹四边形；
图7最小角最大化准则原理图；
图8带判断面示意图，图(a)表示平接接头示意图，图(b)表示角接接头示意图，图(c)表示T型接头示意图；
图9轮廓线条选取示意图，图(a)表示前8组轮廓线示意图，图(b)表示前10组轮廓线示意图；图(c)表示前12组轮廓线示意图；
图10四种三角剖分方法的测试结果；
图11“梳子”多边形四种三角剖分的测试结果；
在图10和图11中，图(a)表示单连通多边形环形追踪算法测试结果，图(b)表示基于凹凸顶点的三角剖分算法测试结果，图(c)表示动态Delaunay三角剖分算法测试结果，图(d)表示本发明测试结果；
图12船舱段的焊缝特征识别测试，图12(a)表示船舱段焊缝案例编号，图12(b)、图12(c)图12(d)、图12(e)图12(f)、图12(g)分别表示焊缝1、2、3、4、5和6特征。
具体实施方式
为使本发明实施例的目的和技术方案更加清楚，下面将结合本发明实施例的附图，对本发明实施例的技术方案进行清楚、完整地描述。显然，所描述的实施例是本发明的一部分实施例，而不是全部的实施例。基于所描述的本发明的实施例，本领域普通技术人员在无需创造性劳动的前提下所获得的所有其他实施例，都属于本发明保护的范围。
如图1所示，本发明的一种焊缝特征识别方法。首先通过角系数法判断多边形的凹凸顶点，基于凹顶点和三角形旋向的Delaunay三角剖分，构造三维模型表面的三角形网格并生成STL文件；然后基于相邻三角面片的法向量夹角，提取出模型的轮廓线及点；最后根据接头空间位置和最小轮廓线距离识别出焊缝特征。具体采用如下步骤：
(1)基于凹顶点和三角形旋向的Delaunay三角剖分；
(2)生成三维模型的STL文件；
(3)提取模型的轮廓线及点；
(4)接头空间位置判断；
(5)焊缝特征识别。
如图2，3，4，5，6和7所示，所述基于凹顶点和三角形旋向的Delaunay三角剖分，包括初始三角网格生成和三角网格局部变换，其中，所述初始三角网格生成的方法，采用如下步骤：
(1)逆时针顺序读入多边形的顶点P 1,P 2,…,P n，并依次存储到数组空间；
(2)通过角系数法判断每个顶点的凹凸性，获取所有凹顶点，且判断方法如下：
如图3所示，在待定顶点P i+1的上、下设置两条距其距离相等的水平直线l 1、l 2，直线l 1或l 2分别与前边P iP i+1、后边P i+1P i+2的交点为A、B。设直线AP i+1、BP i+1的斜率分别为tanα、tanβ，计算角系数差值Δ。若：(ⅰ)交点A、B都在同一条水平直线上且Δ>0，则顶点P i+1为凹顶点；反之，Δ<0，则顶点P i+1为凸顶点。(ⅱ)交点A、B不在同一条水平直线上且Δ>0，则顶点P i+1为凸顶点；反之，Δ<0，则顶点P i+1为凹顶点。其计算如下：
Figure PCTCN2022127298-appb-000001
式中，x a,y a为交点A横纵坐标值；x b,y b为交点B横纵坐标值；x Pi+1,y Pi+1为顶点P i+1横纵坐标值。
此外，若前边P iP i+1或后边P i+1P i+2为水平直线时，则上式1/tanα＝0或1/tanβ＝0。
(3)
Figure PCTCN2022127298-appb-000002
按顺序从数组空间中依次取出三个顶点P j、P j+1、P j+2来构成有向三角形△P j P j+1P j+2，求出其面积S △Pj Pj+1Pj+2进行三角形旋向判断。如图4所示，若计算结果为正，则该三角形为逆时针旋向并转步骤(4)；反之为顺时针旋向并转步骤(5)。若三个顶点位于同一直线上，则也转步骤(5)。其计算如下：
Figure PCTCN2022127298-appb-000003
式中，(x 0,y 0)为顶点P j坐标；(x 1,y 1)为顶点P j+1坐标；(x 2,y 3)为顶点P j+2坐标。
(4)判断三角形内部是否含有凹顶点。若无凹顶点，则输出构成该三角形的三顶点P i、P i+1、P i+2，并将P i+1从数组空间中移除，同时以该三角形的起点P i、终点P i+2和下一个顶点P i+3，构成新有向三角形△P iP i+2P i+3并转步骤(6)；若有凹顶点，则转步骤(5)。判断方法如下：
如图5所示，将待定凹顶点D与三角形顶点P i、P i+1、P i+2相连，则得到∠P iDP i+1、∠P iDP i+2、∠P i+1DP i+2三个角。若点D在三角形ABC内部或边上，即∠ADB+∠ADC+∠BDC＝360°，则三角形内部含有凹顶点；若点D在三角形ABC外部，即∠ADB+∠ADC+∠BDC＜360°，则三角形内部不含凹顶点。
(5)将三角形ΔP iP i+2P i+3的起点P i存回数组空间，取出下一个顶点P i+1，重新构成新有向三角形△P i+1P i+2P i+3并转步骤(6)；
(6)判断数组空间内顶点个数小于等于3？，若是，则结束三角形网格划分，否则获取步骤(4)或步骤(5)中新构成的有向三角形，即△P iP i+2P i+3或△P i+1P i+2P i+3，再转步骤(3)进行旋向判断。
如图6和图7所示，所述三角网格局部变换，是基于最小角最大准则进行的，使得网 格剖分的三角形形态最优，即去除掉初始划分三角形的边界边，对剩余的每一条非边界边按以下步骤处理：
(1)遍历初始三角形网格，找出以非边界边为公共边的两个相邻三角形，并判断由这两个三角形所构成四边形的凹凸性。若为凹四边形，不作任何处理；否则转步骤(2)。四边形凹凸性判断方法如下：
如图6所示，将两相邻三角形的公共边记为直线l 1，将两三角形中不在公共边上的两顶点所连直线记为l 2。若直线l 1与l 2相交，则为凸四边形；反之为凹四边形；
(2)计算两三角形中的六个内角并取出最小角α，再按照图7进行对角线交换，并计算两新三角形中六个内角的最小角β；
(3)比较α、β两角大小。若α≥β，保留对角线交换前两三角形；反之保留对角线交换后两新三角形。
所述生成三维模型的STL文件的方法，主要是通过调用SolidWorks软件API函数，遍历三维模型的每个轮廓面，提取出每个轮廓面上的法向量(指向实体外部)和轮廓顶点坐标信息。将每个轮廓面的顶点按逆时针排列，利用上述三角网格剖分算法，将每次划分输出的三个顶点坐标作为三角面片的三个顶点坐标，三角面片的法向矢量为所在轮廓面的法向量，从而构成一个三角面片的几何信息。通过对整个模型三角剖分即可得到ASCII码文件格式的STL文件。
如图8所示，两接头的待判定面位置关系直接影响接头形式。由于待判定面相对于坡口端面的面积较大，求出每个轮廓面面积并取最大的前四组，即可得到待判定面；然后根据待判定面处法向量(虚线)的平行或垂直关系进行4类接头的空间位置关系判别。
如图9所示，根据10种焊接坡口的轮廓线条数，通常可分为24条、27条和30条三大类。当轮廓线条为24条时，主要有平接接头的I型坡口、V型坡口；角接接头的不开坡口、单边坡口；T型和搭接接头的不开坡口六种类型。图9(a)即属于24条轮廓线类型中平接接头所对应的I型坡口。此类型所涉及焊缝特征识别，主要是先取出前8组长度最短的轮廓线；然后针对该8组轮廓线，计算两两轮廓线的端点间所有距离；最后根据接头空间位置和最小轮廓线距离，进行接头形式和坡口类型的识别。当轮廓线条为27条时，只有角接接头双边坡口和T型接头单边坡口两种。图9(b)即属于27条轮廓线类型中角接接头所对应的双边坡口。此类型所涉及焊缝特征识别，主要是针对最短10组轮廓线，并参照24条轮廓线中的相同处理方式，进行相应接头形式和坡口类型的识别。当轮廓线条为30条时，同样只有平接接头Y型坡口和T型接头双边坡口两种。图9(c)即属于30条轮廓线类型中T型接头所对应的双边坡口。焊缝特征识别主要是针对12组最短轮廓线并参照上述相同方式进行接头形式和坡口类型的识别。
所述提取模型的轮廓线及点的方法是，对STL文件中所有的三角面片两两组合，对组合成的每一对三角面片按如下步骤处理：
(1)判断一对三角面片是否存在两个公共顶点？若是，则相邻；反之不相邻。对于相邻三角面片，则计算出两者的法向量夹角，并转步骤(2)；若不相邻，则取下一对三角面片继续本步骤的相邻判断。其计算如下：
Figure PCTCN2022127298-appb-000004
式中，(x,y,z) 1,2分别为两三角面片的法向量。
(2)根据法向量夹角确定两三角面片的公共线段是否为模型轮廓线。若θ＝0或θ＝180°，则两三角面片处于同一平面内，其两个公共顶点所组成的公共线段不是模型轮廓线；若0<θ<180°，则两三角面片处于不同平面内，公共线段是模型轮廓线的一部分；
(3)输出步骤(2)中所有位于不同平面内的三角面片公共线段，即可提取到模型的全部轮廓；
(4)统计各三角面片顶点在全部轮廓线条中出现的次数n。若n≥3，则该点为模型端点，进而获得所有轮廓顶点；
所述接头空间位置判断，主要是基于带判定面处法向量平行、垂直和方向相反的次数进行判断，具体采用如下步骤：
(1)基于海伦公式计算STL中三角面片面积，并求出每个轮廓面面积。取面积最大的前四组，并获取对应轮廓面的法向量。海伦公式如下：
Figure PCTCN2022127298-appb-000005
其中：p＝(a+b+c)/2为半周长；a、b、c为边长。
(2)对步骤(1)获取的法向量进行两两叉积并求出向量积的模长。若模长为0，则两个法向量平行。统计法向量出现平行的次数P a。其计算如下：
Figure PCTCN2022127298-appb-000006
式中，(l,m,n)为轮廓面法向量
Figure PCTCN2022127298-appb-000007
坐标；(o,p,q)为轮廓面法向量
Figure PCTCN2022127298-appb-000008
坐标。
(3)对步骤(1)获取的法向量进行两两点积并求出数量积。若积为0，则两个法向量垂直。统计法向量出现垂直的次数V t。其计算如下：
Figure PCTCN2022127298-appb-000009
式中，(l,m,n)为轮廓面法向量
Figure PCTCN2022127298-appb-000010
坐标；(o,p,q)为轮廓面法向量
Figure PCTCN2022127298-appb-000011
坐标。
(4)对步骤(1)中获取的法向量进行两两方向判断，并统计出方向相反的次数O p。
(5)根据P a,V t,O p的值判断接头空间位置关系。若P a＝6且O p＝4，则两接头处于平行状态；若V t＝4且O p＝2，则两接头处于垂直状态。
所述焊缝特征识别主要包括接头形式识别和坡口类型识别，所述接头形式识别，主要包括平接、搭接、角接和T型接头4类接头形式的识别，采用如下步骤：
(1)根据轮廓线总条数获取各类最短轮廓线组。若为24条，则输出前8组长度最短的轮廓线组；若为27条，则输出前10组最短轮廓线组；若为30条，则输出前12组最短轮廓线组；
(2)取次短长度的轮廓线作为距离阈值。针对步骤(1)中输出的最短轮廓线组，计算两两轮廓线的端点间所有距离，去除轮廓线本身长度后，将最小距离与距离阈值对比。若最小距离大于距离阈值，则称轮廓线距离为大间距；反之称为小间距；
(3)根据接头空间位置和轮廓线距离进行接头形式识别。若接头处于平行状态，当 轮廓线距离为小间距时，则为平接接头；反之为搭接接头。若接头处于垂直状态，当轮廓线距离为小间距时，则为角接接头；反之为T型接头。
所述坡口类型识别，主要包括平接接头的I型、V型和Y型坡口；角接和T型接头的不开、单边和双边坡口和搭接接头的不开坡口10种坡口类型的识别，采用如下步骤：
(1)根据轮廓线总条数和接头形式进行坡口类型识别。若轮廓线总条数为24条，当接头形式为平接或角接时，转步骤(2)；当为搭接或T型接头时，则接头都为不开坡口。若轮廓线总条数为27条，转步骤(3)。若轮廓线总条数为30条，转步骤(4)；
(2)根据两条最小距离轮廓线的夹角α进行坡口识别。若接头形式为平接，当α＝0时，则为I型坡口；反之为V型坡口。若接头形式为角接，当α＝90°时，则为不开坡口；反之为单边坡口；
(3)根据轮廓线距离进行坡口识别。若为小间距，则为角接接头的双边坡口；反之为T型接头的单边坡口；
(4)根据接头形式和轮廓线距离进行坡口识别。若为平接接头且小间距，则为平接接头的Y型坡口；若为T型接头且大间距，则为T型接头的双边坡口。
为了验证本发明一种船舶焊缝特征识别方法的有效性和稳定性，对“梳子”等多边形进行了剖分测试，并与单连通多边形环形追踪算法(单连通追踪法) [1]、基于凹凸顶点的三角剖分算法(凹凸顶点剖分法) [2]和动态Delaunay三角剖分算法(动态Delaunay法) [3]进行了比较。四种方法两类多边形的剖分结果如图10和图11所示，剖分对比的性能指标采用网格平均质量系数q M和网格关联质量系数q J。
附对比文献：
[1]周知.三角剖分算法研究[D].哈尔滨:哈尔滨理工大学,2007.
[2]李岚.基于凹凸顶点判定的简单多边形区域的三角剖分[J].中国计量学院学报,2001(1):166-168.
[3]罗小华,付文超,管培祥,等.简单多边形的动态Delaunay三角剖分算法[J].暨南大学学报(自然科学与医学版),2011,32(1):26-30,5.
由图10和图11可以看出，单连通追踪法中畸形三角形数量最多，网格形态较差，而本发明算法畸形三角形数量较少与其相比有着明显优化。对“梳子”多边形划分时，动态Delaunay法无法实现剖分，而本发明算法剖分测试正常且在畸形三角形数量的控制上也优于其余两种方法，剖分质量最好。
表一为网格质量分析结果，由表可以看出，无论是网格平均质量系数还是网格关联质量系数，本发明算法的剖分质量都优于其他三种方法，且针对图10多边形分别平均增加了12.49％和12.78％，针对图11多边形分别平均增加了11.62％和11.74％，由此进一步说明本发明算法有效降低了网格畸形程度，使得全体网格更接近于正三角形，保证了较好网格质量的产生。
表一：
Figure PCTCN2022127298-appb-000012
为了进一步验证本发明算法的有效性，针对图12中船舱段上的6处焊缝进行了特征识别测试，并与基于焊接特征的焊接工件库建模(特征库设计法) [4]和基于SolidWorks的遍历特征树技术(遍历特征法) [5]进行了比较。
附对比文献：
[4]应灿.焊接机器人工作站协同运动的轨迹优化研究[D].广州:华南理工大学,2013.
[5]张泽南,仲梁维.基于SolidWorks的自动焊缝标注的二次开发[J].软件工程,2020,23(5):17-20.
表二为三种焊缝特征识别方法的测试结果，由表可以看出，遍历特征法仅能识别出平接接头和T型接头两种形式，且对坡口类型无法识别，识别效果较差。特征库设计法虽然与本发明算法一样，都能准确识别出6处焊缝特征，但在接头和坡口识别的平均运行时间上均高于本发明算法，执行效率低，此外，特征库设计法需要先对船舱段上的所有焊缝，进行特征造型设计并建立数据库，使得开发工作量大大增加，且当数据库规模较大在匹配焊缝特征时可能出现耗时较长的情况。而本发明算法在接头和坡口识别的平均运行时间上更短，执行效率更高，且只需获取模型焊缝的STL文件，前期处理量较小，同时识别的接头形式、坡口类型多，准确率高。
表三、表四分别为本发明算法的接头和坡口详细测试结果，由表可以看出，本发明算法能快速统计出待判定面处法向量平行、垂直、相反的次数，精确获取最小轮廓线距离、距离阈值、轮廓线总条数和两条最小距离轮廓线的夹角，并根据关键参数作出正确判断和对比，准确识别出6处焊缝的接头形式和坡口类型，即平接接头I型坡口、平接接头V型坡口、T型接头不开坡口、角接接头不开坡口、角接接头双边坡口和T型接头双边坡口，从而实现焊缝特征的精确识别，执行效率高，验证了本发明算法的有效性。
表二：
Figure PCTCN2022127298-appb-000013
表三：
Figure PCTCN2022127298-appb-000014
表四：
Figure PCTCN2022127298-appb-000015
以上实施例仅用以说明本发明的技术方案而非对其限制，尽管参照上述实施例对本发明进行了详细的说明，所属领域的普通技术人员应当理解：依然可以对本发明的具体实施方式进行修改或者等同替换，而未脱离本发明精神和范围的任何修改或者等同替换，其均应涵盖在本发明的权利要求保护范围之内。

Claims (7)

  1. 一种船舶焊缝特征识别方法，其特征在于，包括如下步骤：
    (1)基于凹顶点和三角形旋向的Delaunay三角剖分；
    所述基于凹顶点和三角形旋向的Delaunay三角剖分，包括初始三角网格生成和三角网格局部变换；
    (2)生成三维模型的STL文件；
    所述生成三维模型的STL文件，主要是通过遍历三维模型的每个轮廓面信息，利用上述Delaunay三角网格剖分算法，获取三角面片的几何信息；通过对整个模型三角剖分即得到ASCII码文件格式的STL文件；
    (3)提取模型的轮廓线及点；
    (4)接头空间位置判断；
    (5)焊缝特征识别；
    所述焊缝特征识别主要包括接头形式识别和坡口类型识别。
  2. 根据权利要求1所述的一种船舶焊缝特征识别方法，其特征在于，步骤(1)中所述初始三角网格生成的具体内容和步骤如下：
    1)逆时针顺序读入多边形顶点，并依次存储到数组空间；
    2)通过角系数法判断每个顶点的凹凸性，获取所有凹顶点；
    3)按顺序从数组空间中依次取出三个顶点来构成有向三角形，求出其面积进行三角形旋向判断；当计算结果为正，则该三角形为逆时针旋向并转步骤4)；反之为顺时针旋向并转步骤5)；当三个顶点位于同一直线上，则也转步骤5)；
    4)判断三角形内部是否含有凹顶点；当无凹顶点，则输出构成该三角形的三顶点P i、P i+1、P i+2，并将P i+1从数组空间中移除，同时以该三角形的起点P i、终点P i+2和下一个顶点P i+3，构成新有向三角形△P i P i+2 P i+3并转步骤6)；当有凹顶点，则转步骤5)；
    5)将三角形△P i P i+2 P i+3的起点P i存回数组空间，取出下一个顶点P i+1，重新构成新有向三角形△P i+1 P i+2 P i+3并转步骤6)；
    6)判断数组空间内顶点个数小于等于3？当是，则结束三角形网格划分，否则获取步骤4)或步骤5)中新构成的有向三角形，再转步骤3)进行旋向判断。
  3. 根据权利要求1所述的一种船舶焊缝特征识别方法，其特征在于，步骤(1)中所述三角网格局部变换的方法是，去除掉初始划分三角形的边界边，对剩余的每一条非边界边按以下方法和步骤处理：
    (a)遍历初始三角形网格，找出以非边界边为公共边的两个相邻三角形，并判断由这两个三角形所构成四边形的凹凸性；当为凹四边形，则不作任何处理；否则转步骤(b)；
    (b)计算两三角形中的六个内角并取出最小角α，再进行对角线交换，并计算两新三角形中六个内角的最小角β；
    (c)比较α、β两角大小；当α≥β，则保留对角线交换前两三角形；反之保留对角线交换后两新三角形。
  4. 根据权利要求1所述的一种船舶焊缝特征识别方法，其特征在于，步骤(3)所述提取模型的轮廓线及点的方法是，对STL文件中所有的三角面片两两组合，将组合成的每一对三角面片按如下步骤进行处理：
    (d)判断一对三角面片是否存在两个公共顶点？当是，则相邻；反之不相邻；对于相邻三 角面片，则计算出两者的法向量夹角，并转步骤(e)；当不相邻，则取下一对三角面片继续本步骤的相邻判断；
    (e)根据法向量夹角确定两三角面片的公共线段是否为模型轮廓线；当θ＝0或θ＝180°，则两个公共顶点所组成的公共线段不是模型轮廓线；当0<θ<180°，则公共线段是模型轮廓线的一部分；
    (f)输出步骤(e)中所有位于不同平面内的三角面片公共线段，即提取出模型的全部轮廓；
    (g)统计各三角面片顶点在全部轮廓线条中出现的次数n；当n≥3，则该点为模型端点，进而获得所有轮廓顶点。
  5. 根据权利要求1所述的一种船舶焊缝特征识别方法，其特征在于，步骤(4)所述接头空间位置判断的具体方法，采用如下步骤：
    (h)基于海伦公式计算STL中三角面片面积，并求出每个轮廓面面积；取面积最大的前四组，并获取对应轮廓面的法向量；
    (i)对步骤(h)获取的法向量进行两两叉积并求出向量积的模长；若模长为0，则两个法向量平行；统计法向量出现平行的次数P a；
    (j)对步骤(h)获取的法向量进行两两点积并求出数量积；当积为0，则两个法向量垂直；统计法向量出现垂直的次数V t；
    (k)对步骤(h)中获取的法向量进行两两方向判断，并统计出方向相反的次数O p；
    (l)根据P a,V t,O p的值判断接头空间位置关系；当P a＝6且O p＝4，则两接头处于平行状态；当V t＝4且O p＝2，则两接头处于垂直状态。
  6. 根据权利要求1所述的一种船舶焊缝特征识别方法，其特征在于，步骤(5)中所述接头形式识别的具体方法，采用如下步骤：
    (m)根据轮廓线总条数获取各类最短轮廓线组；当为24条，则输出前8组长度最短的轮廓线组；当为27条，则输出前10组最短轮廓线组；当为30条，则输出前12组最短轮廓线组；
    (n)取次短长度的轮廓线作为距离阈值；针对步骤(m)中输出的最短轮廓线组，计算两两轮廓线的端点间所有距离，去除轮廓线本身长度后，将最小距离与距离阈值对比；当最小距离大于距离阈值，则称轮廓线距离为大间距；反之称为小间距；
    (o)根据接头空间位置和轮廓线距离进行接头形式识别；当接头处于平行状态，和轮廓线距离为小间距时，则为平接接头；当接头处于平行状态，和轮廓线距离为大间距时，为搭接接头；当接头处于垂直状态，和轮廓线距离为小间距时，则为角接接头；当接头处于垂直状态，和轮廓线距离为大间距时，为T型接头。
  7. 根据权利要求1所述的一种船舶焊缝特征识别方法，其特征在于，步骤(5)中所述坡口类型识别的具体方法，采用如下步骤：
    (p)根据轮廓线总条数和接头形式进行坡口类型识别；当轮廓线总条数为24条，和接头形式为平接或角接时，转步骤(q)；当为搭接或T型接头时，则接头都为不开坡口；若轮廓线总条数为27条，转步骤(s)；若轮廓线总条数为30条，转步骤(t)；
    (q)根据两条最小距离轮廓线的夹角α进行坡口识别；当接头形式为平接，和α＝0时，则为平接接头的I型坡口；当接头形式为平接，和α≠0时，则为平接接头的V型坡口；当接头形式为角接，和α＝90°时，则为角接接头的不开坡口；当接头形式为角接，和α≠90°时，则为角 接接头的单边坡口；
    (s)根据轮廓线距离进行坡口识别；当为小间距，则为角接接头的双边坡口；反之为T型接头的单边坡口；
    (t)根据接头形式和轮廓线距离进行坡口识别；当为平接接头且小间距，则为平接接头的Y型坡口；当为T型接头且大间距，则为T型接头的双边坡口。
PCT/CN2022/127298 2021-12-07 2022-10-25 一种船舶焊缝特征识别方法 WO2023103621A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202111483243.1A CN114332475B (zh) 2021-12-07 2021-12-07 一种船舶焊缝特征识别方法
CN202111483243.1 2021-12-07

Publications (1)

Publication Number Publication Date
WO2023103621A1 true WO2023103621A1 (zh) 2023-06-15

Family

ID=81048057

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/127298 WO2023103621A1 (zh) 2021-12-07 2022-10-25 一种船舶焊缝特征识别方法

Country Status (2)

Country Link
CN (1) CN114332475B (zh)
WO (1) WO2023103621A1 (zh)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116532800A (zh) * 2023-07-06 2023-08-04 武汉创恒激光智能装备有限公司 一种汽车节气门阀片组件激光焊接装置
CN116956465A (zh) * 2023-07-26 2023-10-27 中基科技(武汉)有限公司 一种基于xml的船体结构模型构建方法、装置、设备及介质
CN117934616A (zh) * 2024-03-21 2024-04-26 深圳前海瑞集科技有限公司 船舶工件的焊缝确定方法和装置

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114332475B (zh) * 2021-12-07 2024-09-06 江苏科技大学 一种船舶焊缝特征识别方法
CN114769966B (zh) * 2022-05-13 2023-12-26 中车青岛四方机车车辆股份有限公司 车体断面焊缝识别方法及系统
CN117400267B (zh) * 2023-12-12 2024-03-29 上海船舶工艺研究所(中国船舶集团有限公司第十一研究所) 机器人焊接路径生成方法、装置、计算机设备和存储介质

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2012234255A (ja) * 2011-04-28 2012-11-29 Suzuki Motor Corp 画像処理装置及び画像処理方法
CN111730256A (zh) * 2020-08-31 2020-10-02 季华实验室 基于数字模型的焊缝识别方法、装置、存储介质及终端
CN111783260A (zh) * 2020-08-14 2020-10-16 广东科学技术职业学院 一种面向stl文件的截面轮廓线段连接方法
CN113012259A (zh) * 2021-03-23 2021-06-22 上海葛兰岱尔网络科技有限公司 一种基于三角剖分算法对凹多边形填充处理的方法
CN114332475A (zh) * 2021-12-07 2022-04-12 江苏科技大学 一种船舶焊缝特征识别方法

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR102700117B1 (ko) * 2017-11-06 2024-08-28 헴펠 에이/에스 해양 선박의 유체 동적 프로파일을 개선하는 방법, 개선된 유체 동적 프로파일을 갖는 해양 선박 및 유체 동적 프로파일을 개선하기 위한 코팅 시스템
CN111709083B (zh) * 2020-04-30 2023-03-07 广州文冲船厂有限责任公司 一种焊接方位的判别方法
CN112288707B (zh) * 2020-10-27 2024-08-16 南京神源生智能科技有限公司 一种基于特征点识别的机器人焊缝打磨算法
CN112819806B (zh) * 2021-02-23 2024-05-28 江苏科技大学 一种基于深度卷积神经网络模型的船舶焊缝缺陷检测方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2012234255A (ja) * 2011-04-28 2012-11-29 Suzuki Motor Corp 画像処理装置及び画像処理方法
CN111783260A (zh) * 2020-08-14 2020-10-16 广东科学技术职业学院 一种面向stl文件的截面轮廓线段连接方法
CN111730256A (zh) * 2020-08-31 2020-10-02 季华实验室 基于数字模型的焊缝识别方法、装置、存储介质及终端
CN113012259A (zh) * 2021-03-23 2021-06-22 上海葛兰岱尔网络科技有限公司 一种基于三角剖分算法对凹多边形填充处理的方法
CN114332475A (zh) * 2021-12-07 2022-04-12 江苏科技大学 一种船舶焊缝特征识别方法

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116532800A (zh) * 2023-07-06 2023-08-04 武汉创恒激光智能装备有限公司 一种汽车节气门阀片组件激光焊接装置
CN116532800B (zh) * 2023-07-06 2023-09-22 武汉创恒激光智能装备有限公司 一种汽车节气门阀片组件激光焊接装置
CN116956465A (zh) * 2023-07-26 2023-10-27 中基科技(武汉)有限公司 一种基于xml的船体结构模型构建方法、装置、设备及介质
CN117934616A (zh) * 2024-03-21 2024-04-26 深圳前海瑞集科技有限公司 船舶工件的焊缝确定方法和装置
CN117934616B (zh) * 2024-03-21 2024-05-28 深圳前海瑞集科技有限公司 船舶工件的焊缝确定方法和装置

Also Published As

Publication number Publication date
CN114332475B (zh) 2024-09-06
CN114332475A (zh) 2022-04-12

Similar Documents

Publication Publication Date Title
WO2023103621A1 (zh) 一种船舶焊缝特征识别方法
US20210173966A1 (en) Method for constructing curve of robot processing path of part with small curvature based on point cloud boundary
CN107123164B (zh) 保持锐利特征的三维重建方法及系统
CN109919984A (zh) 一种基于局部特征描述子的点云自动配准方法
CN105469404A (zh) 一种基于三维点云数据的旋转体拟合方法及装置
CN107330142B (zh) 一种基于stl模型的在机检测的测点法矢估算方法
CN114240871B (zh) 一种面向工件成型过程中轮廓检测的点云数据处理方法
CN112819883B (zh) 一种规则对象检测及定位方法
CN110415331B (zh) 一种基于点云数据的轮廓或孔洞快速检测判别和孔洞修补方法
CN115018249A (zh) 一种基于激光扫描技术的地铁站施工质量评价方法
CN115222883A (zh) 一种基于地基LiDAR点云的电力杆塔重建方法
CN116245928A (zh) 一种基于双目立体匹配的三维重建方法
CN110363801B (zh) 工件实物与工件三维cad模型的对应点匹配方法
Masuda et al. As-built 3D modeling of large facilities based on interactive feature editing
CN117444968A (zh) 一种大型筒件局部特征机器人扫描测量路径规划方法
Zhang et al. Feature extraction and robot path planning method in 3D vision-guided welding for multi-blade wheel structures
Zhang et al. A Feature Extraction Approach Over Workpiece Point Clouds for Robotic Welding
CN116957963A (zh) 基于分层k领域的输电线路激光点云数据噪声处理方法
CN112818428B (zh) 一种用于cad模型面结构光全自动扫描路径规划方法
CN111583098B (zh) 一种基于序列图像的线段聚类及拟合方法和系统
WO2023082533A1 (zh) 一种船舶焊缝特征参数提取方法
Ren et al. A 3D point cloud registration algorithm based on feature points
CN114549740B (zh) 岩体裂隙三维模型构建及特征参数智能提取方法及应用
Chen et al. Denoising algorithm for bilateral filtered point cloud based on variance threshold
CN114722938B (zh) 一种混凝土梁点云的Voronoi疏密度分类逆向重建方法

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22903045

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE