基于视觉的线材自动挂牌机器人系统构建与试验  

Construction and Experiment of a Vision Based Automatic Wire Tagging Robot System

作　　者：侯雨雷[1] 万宁静 范德斌 韩延伟 李岳 蔡庆明 曾达幸 HOU Yulei;WAN Ningjing;FAN Debin;HAN Yanwei;LI Yue;CAI Qingming;ZENG Daxing(School of Mechanical Engineering,Yanshan University,Qinhuangdao Hebei 066004,China;School of Mechanical Engineering,Dongguan University of Technology,Dongguan Guangdong 523808,China)

机构地区：[1]燕山大学机械工程学院,河北秦皇岛066004 [2]东莞理工学院机械工程学院,广东东莞523808

出　　处：《机械设计与研究》2024年第5期131-137,149,共8页Machine Design And Research

基　　金：广东省基础与应用基础研究基金自然科学基金(2023A1515012103);广东高校科研平台和项目-创新团队项目(自科)(2022KCXTD033);广东省重点建设学科科研能力提升项目(2021ZDJS084)。

摘　　要：目前,钢厂仍多采用人工形式进行线材挂牌作业,效率低下,劳动强度高。针对此,搭建线材数据采集环境,对3D点云数据进行预处理,开发基于降维的感兴趣区域定位算法,初步确定线材挂牌区域。提出“切四刀”方法,将大数据切割得到关键三维区域,根据捆线的几何特征,精准定位线材挂牌位置的三维坐标。利用2D目标检测算法对标牌位置进行识别,确定挂牌位置情况。进行现场整机试验,结果表明所构建的基于视觉的线材自动挂牌机器人系统挂牌准确率达到97%以上,满足现场应用需求。研究工作实现了线材自动挂牌,对类似作业自动化水平的提升具有指导意义。At present,steel mills still use manual methods for wire labeling operations,which is inefficient and labor-intensive.In response to this,a wire data collection environment is established,3D point cloud data is preprocessed,a dimensionality reduction-based region of interest localization algorithm is developed,and the wire listing area is preliminarily determined.The“four knife”method is proposed to cut big data into key three-dimensional areas,and accurately locate the three-dimensional coordinates of the wire hanging position based on the geometric characteristics of the bundling line.2D object detection algorithm is used to identify the position of the sign and determine the listing position.The on-site experiment is conducted,and the results show that the visual based automatic wire tagging robot system constructed can achieve a tagging accuracy of over 97%,meeting the on-site application requirements.The research work has achieved automatic wire tagging,which has guiding significance for improving the automation level of similar operations.

关 键 词：线材挂牌 3D视觉 点云处理 目标检测 识别定位 

分 类 号：TH242[机械工程—机械制造及自动化]