基于3D结构光视觉引导的件烟堆垛识别与定位系统设计  

Design of Cigarette Stack Recognition and Positioning System Based on 3D Structured Light Visual Guidance

作　　者：谢娟 李逸云 向光禄 乐科友 宋杰 XIE Juan;LI Yiyun;XIANG Guanglu;LE Keyou;SONG Jie(Chongqing Tobacco Company Wanzhou Branch,Chongqing 404100,China;Chongqing Tobacco Company Logistics Branch,Chongqing 400000,China)

机构地区：[1]重庆市烟草公司万州分公司,重庆404100 [2]重庆市烟草公司物流分公司,重庆400000

出　　处：《移动信息》2025年第3期295-297,共3页Mobile Information

摘　　要：文中在分析当前烟草卸货物流现状的基础上,针对运输车辆规格随机变更引起的件烟随机堆垛,导致件烟间的空隙不易识别、吸盘操作效率低下的问题,设计了一种视觉引导的件烟识别和定位系统。该系统基于深度学习算法pointNet++点云分割,采用自适应最大外接矩形法,并结合目标检测交并比算法,提高了对抓取区域的识别精度,再借助几何交并比对抓取区域进行自由化,以交互式定位系统的多次抓取策略提升了整个专用卸货系统的工作效率及可靠性。Based on the analysis of the current status of tobacco unloading logistics,this paper designs a visual-guided cigarette recognition and positioning system for the problems of random stacking of pieces and cigarettes caused by random changes in transportation vehicle specifications,which leads to difficult identification of gaps between pieces and low operating efficiency of suction cups.The system is based on deep learning algorithm pointNet++point cloud segmentation,adopts adaptive maximum external rectangle method,and combines object detection intersection ratio algorithm to improve the recognition accuracy of the grasping area.Then,the geometric intersection and comparison are used to liberalize the grasping area.The working efficiency and reliability of the entire dedicated unloading system are improved by the multiple grasping strategy of the interactive positioning system.

关 键 词：件烟堆垛 结构光 视觉引导 识别及定位 

分 类 号：TP311.5[自动化与计算机技术—计算机软件与理论]