融合2D-3D视觉的工业机器人抓取系统设计  

Design of Industrial Robot Grasping System with 2D-3D Vision Integration

作　　者：李跃明 杨牧 LI Yueming;YANG Mu(Huangshi Xinxing Pipe Industry Co.,Ltd.,Huangshi Hubei 435000;Safety and Environmental Protection Department of Handan Xinxing Cast Pipe Co.,Ltd.,Handan Hebei 056300)

机构地区：[1]黄石新兴管业有限公司,湖北黄石435000 [2]河北省邯郸市新兴铸管股份有限公司安全环保部,河北邯郸056300

出　　处：《软件》2025年第3期156-158,共3页Software

摘　　要：在工业制造持续向柔性化和智能化演进的背景下,传统机器人在处理无序堆叠目标和复杂作业场景方面面临识别精度不足、抓取效率低等问题。基于此,本文设计了一种融合2D-3D视觉的工业机器人抓取系统,构建了“视觉感知—数据融合—抓取规划—机器人执行”分层架构,融合RGB图像与结构光点云信息,实现了高精度目标位姿感知与自适应路径规划。系统深入集成了YOLOv5检测算法、ICP配准与RRT*动态避障策略,显著提高了目标抓取稳定性,减少了人工干预与误差率,提升了在复杂工业场景下的抓取效率与系统鲁棒性,为柔性生产提供了技术保障。In the context of industrial manufacturing's shift towards flexibility and intelligence,traditional robots struggle with low recognition accuracy and grasping efficiency in complex scenarios.To address this,a 2D-3D vision-integrated grasping system for industrial robots is proposed.It features a hierarchical architecture combining visual perception,data fusion,grasping planning,and robot execution.By integrating RGB images and structured light point clouds,it achieves precise target pose perception and adaptive path planning.The system integrates YOLOv5 detection,ICP registration,and RRT*dynamic obstacle avoidance,enhancing grasping stability,reducing manual intervention and errors,and improving efficiency and robustness in complex industrial settings.

关 键 词：2D-3D视觉 工业机器人 抓取系统设计 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]