机电工程技术2025，Vol.54Issue(19)：19-24,64,7.DOI:10.3969/j.issn.1009-9492.2025.00052
基于双向的修剪ICP点云配准算法
Bidirectional Trimmed ICP Point Cloud Registration Algorithm
摘要
Abstract
In the manufacturing process of large castings,surface defects such as flash and burrs are common,which need to be ground and cleaned to meet application requirements.Based on a 3D vision-guided robotic processing method,the casting point cloud model is reconstructed,analyzed,and compared to locate the processing trajectory,thereby guiding the robot to complete the grinding task.Multi-view scanned point clouds often have misalignment and overlap,requiring precise registration.A bidirectional trimming ICP point cloud registration model(Bi-TrICP)is proposed,which adaptively iterates to solve the forward and backward point cloud registration overlap rate,effectively eliminating misaligned points and ensuring the algorithms generalization.A 3D vision positioning platform is then constructed,with a rotating platform,industrial robot,and 3D camera working together to collect the casting point cloud data.Multi-view point cloud registration software is used to stitch the hardware point clouds.Experimental results show that the method successfully reconstructs the global model of the bathroom hardware with precise positioning.The reconstructed hardware components show no significant misalignment,and the positioning error is around 0.73 mm,meeting the requirements for subsequent industrial processing.关键词
大型铸件/点云配准/视觉定位Key words
large castings/point cloud registration/visual positioning分类
矿业与冶金引用本文复制引用
吴立华,甘汤键,吴立坤,温涛,申启访,陈泓宇..基于双向的修剪ICP点云配准算法[J].机电工程技术,2025,54(19):19-24,64,7.基金项目
广东省普通高校青年创新人才类项目(2024KQNCX122) （2024KQNCX122）
广东省普通高校重点领域专项(2024ZDZX3044) （2024ZDZX3044）
