基于PointNet++的焊装夹具零件识别  

PointNet++Based Welding Fixture Parts Identification

作　　者：徐华 陶长城 乐鑫淼 XU Hua;TAO Changcheng;YUE Xinmiao(School of Materials Science and Engineering,Hubei University of Automotive Technology,Shiyan 442000,China)

机构地区：[1]湖北汽车工业学院材料科学与工程学院,十堰442000

出　　处：《组合机床与自动化加工技术》2024年第4期141-144,共4页Modular Machine Tool & Automatic Manufacturing Technique

摘　　要：焊装夹具是汽车白车身焊接生产线中重要的组成部分,有效的管理和归纳焊装夹具零件设计数模能够显著提高设计效率。将原始设计数模离散为点云,利用点云数据和PointNet++深度学习网络探讨了一种焊装夹具零件智能分类方法,并对比各模型的分类精度,选取运行效率和精度最高的单尺度分组(SSG)模型完成焊装夹具零件的分类。训练结果表明,该方法在验证集上的准确率为97.5%,型块、连接块、定位销、销座、支座的验证集类内准确率分别为92.5%、97.5%、100%、97.5%和100%。这些结果表明该方法具有较高的识别精度,能够满足焊装夹具零件分类的精度要求。Welding fixture is an important part in the welding production line of automobile body in white.Effective management and induction of welding fixture parts design modulus can significantly improve the design efficiency.In this paper,the original design model is discretized into a point cloud,and an intelligent classification method of welding fixture parts is discussed by using the point cloud data and Pointnet++deep learning network.By comparing the classification accuracy of each model,the single scale grouping(SSG)model with the highest operating efficiency and accuracy is selected to complete the classification of welding fixture parts.The training results show that the accuracy of the proposed method on the verification set is 97.5%,and the accuracy of the verification set of the type block,the connection block,the positioning pin,the pin seat and the support are 92.5%,97.5%,100%,97.5%and 100%,respectively.These results show that the proposed method has high recognition accuracy and can meet the accuracy requirements of welding fixture parts classification.

关 键 词：焊装夹具 三维点云 分类 PointNet++ 

分 类 号：TH161[机械工程—机械制造及自动化] TG506[金属学及工艺—金属切削加工及机床]