基于点云深度学习的散乱堆叠轴承圈位姿检测  

Pose Estimation of Scattered Stacked Bearing Rings Based on Point Cloud Deep Learning

作　　者：麦海锋 姚锡凡[1] MAI Haifeng;YAO Xifan(School of Mechanical and Automotive Engineering,South China University of Technology,Guangzhou 510641,China)

机构地区：[1]华南理工大学机械与汽车工程学院,广州510641

出　　处：《组合机床与自动化加工技术》2023年第11期56-59,64,共5页Modular Machine Tool & Automatic Manufacturing Technique

基　　金：广东省基础与应用基础研究基金项目(2021A1515010506,2022A1515010095);企业产学研合作项目(20201257)。

摘　　要：为解决工业应用中散乱堆叠轴承圈的难识别、分割与抓取问题,提出一种分层递进随机下采样算法对采集的点云模型进行下采样操作,并结合提出的基于RGB阈值自动标注算法完成数据集的制作,利用PointNet++网络预测并分割出可抓取轴承圈上表面,进而使用RANSAC算法精确分割出待抓取轴承圈上表面,最后采用防干涉的抓取点选取策略完成待抓取轴承圈的位姿检测。实际场景下三组抓取实验得到的成功率均在98%以上,验证了其有效性。In order to solve the problem of identification,segmentation and grasping of scattered stacked bearing rings in industrial applications,a hierarchical progressive downsampling algorithm is proposed to downsample the collected point cloud model,and combined with the proposed RGB threshold-based automatic labeling algorithm to complete the production of data sets.The PointNet++network is used to predict that the upper surface of the grabable bearing rings.Then the RANSAC algorithm is used to accurately segment the upper surface of the bearing ring.At last,the anti-interference grasping point selection strategy is used to complete the attitude estimation of the bearing ring.In the actual scenario,the success rates of the three groups of grasping experiments are above 98%,which verifies the effectiveness of the proposed method.

关 键 词：工业零件 点云深度学习 位姿检测 机械臂抓取 

分 类 号：TH133.3[机械工程—机械制造及自动化] TG659[金属学及工艺—金属切削加工及机床]