船舶小组立焊接工位特征提取和检测  

Feature Extraction and Detection of Ship Small Assembly in Welding Position

作　　者：赵超奇 习俊通[1] ZHAO Chao-qi;XI Jun-tong(Institute of Intelligent Manufacturing and Information Engineering,Shanghai Jiaotong University,Shanghai 200240,China)

机构地区：[1]上海交通大学智能制造与信息工程研究所,上海200240

出　　处：《组合机床与自动化加工技术》2021年第8期76-80,共5页Modular Machine Tool & Automatic Manufacturing Technique

基　　金：工信部船舶智能制造专项资助项目([2016]545号);上海市科学技术委员会资助项目(18511107500)。

摘　　要：为实现无人工干预的船舶小组立工件的智能焊接,采用线激光设备采集小组立工件表面点云,并制定针对性的算法对海量三维点云进行特征提取和检测。根据小组立流水线焊接工位现有条件和需求,制定基于二轴位移机构的线激光扫描式测量方案。针对海量线激光点云,从算法复杂度和准确性两方面,设计基于随机采样的底板提取算法,分割底板和加强筋。通过角点提取算法计算底板角点坐标和关键尺寸,对加强筋进行聚类分割和基于线条点云聚类阈值的噪点去除,并通过对比焊接前后的加强筋点云评估焊接质量。通过现场测量实验,验证了测量方案和算法的可靠性,表明机器视觉的方案可以为船舶智能制造提供技术基础。In order to realize the intelligent welding of ship small assembly without manual intervention,the line laser equipment is used to collect the point cloud on the surface of the small assembly,and targeted algorithms are developed for feature extraction and detection of massive 3D point cloud.A line laser scanning measurement scheme based on two-axis displacement mechanism was built.Low complexity and high accuracy algorithms are proposed to process massive point cloud.Random Sample Consensus algorithm is used to extract the baseboard and the stiffeners.Corner extraction algorithm is proposed to extract corners and calculate the corresponding size.Euclidean clustering segmentation algorithm is used to separate different stiffeners.Noise removal based on line point cloud clustering threshold is proposed.The welding quality of stiffeners is evaluated by comparing the point cloud before and after welding.The reliability of the measurement scheme and algorithm in this paper is verified by experiments,which shows the methodology of machine vision can provide technical basis for ship intelligent manufacturing.

关 键 词：小组立 随机采样一致算法 聚类分割 线激光 特征识别 

分 类 号：TH162[机械工程—机械制造及自动化] TG506[金属学及工艺—金属切削加工及机床]