高精度视觉识别技术在摘钩机器人中的应用  

Application of High-Precision Visual Recognition Technology in Hook Removal Robots

作　　者：贾军祥 蔡茁 杜宗达 曹锦利 樊喜山 朱立学 徐公庆 陈禹桥 陈佐玉 JIA Junxiang;CAI Zhuo;DU Zongda;CAO Jinli;FAN Xishan;ZHU Lixue;XU Gongqing;CHEN Yuqiao;CHEN Zuoyu(Liaoning Datang International Huludao Thermoelectric Co.,Ltd.,Huludao,Liaoning Province,125000 China)

机构地区：[1]辽宁大唐国际葫芦岛热电有限责任公司,辽宁葫芦岛125000

出　　处：《大众科学》2024年第23期4-6,共3页China Public Science

摘　　要：研究了高精度视觉识别技术在摘钩机器人中的应用,针对现代工业自动化中摘钩机器人在高风险、高强度作业环境中的需求,提出了一种结合视觉系统与深度学习算法的解决方案。研究对象为摘钩机器人系统,包括机械系统、视觉系统、控制系统和智能决策模块。通过使用高精度工业相机、3D激光扫描仪和深度学习模型,提升机器人在目标检测、特征匹配、3D空间重建及姿态估计中的精度与响应速度。采用了YOLOv5s网络进行目标检测、PointNet++网络用于3D点云分割,并通过改进算法提升检测精度和实时性。结论表明:基于高精度视觉识别技术的摘钩机器人能够显著提高自动化操作的精度与效率,具有良好的环境适应性与稳定性。This paper studies the application of high-precision visual recognition technology in hook-picking robot,and proposes a solution combining visual system and deep learning algorithm.The research object is the hook robot system,including mechanical system,visual system,control system and intelligent decision module.By using high-precision industrial cameras,3D laser scanner and deep learning models,the accuracy and response speed of the robot in target detection,feature matching,3D spatial reconstruction and attitude estimation are improved.The YOLOv5s network is used for target detection,the PointNet++network for 3D point cloud segmentation,and the detection accuracy and real-time are improved by improving the algorithm.The conclusion shows that the robot based on high precision visual recognition technology can significantly improve the accuracy and efficiency of automatic operation,and has good environmental adaptability and stability.

关 键 词：摘钩机器人 高精度视觉识别 目标检测 3D 空间重建 

分 类 号：TP242.62[自动化与计算机技术—检测技术与自动化装置]