基于单目视觉的盘类零件位姿检测  

Pose Detection of Disc Parts Based on Monocular Vision

作　　者：张瑞杰 刘凌云[1] 王佳维 姚天胜 李泽亚 ZHANG Rui-jie;LIU Ling-yun;WANG Jia-wei;YAO Tian-sheng;LI Ze-ya(School of Electrical and Information Engineering,Hubei University of Automotive Technology,Hubei 442002,China)

机构地区：[1]湖北汽车工业学院电气与信息工程学院,湖北442002

出　　处：《组合机床与自动化加工技术》2022年第9期64-67,共4页Modular Machine Tool & Automatic Manufacturing Technique

基　　金：湖北省自然科学基金项目(2016CFB401)。

摘　　要：针对汽车零部件自动装配技术中的位姿估计,提出以同一平面不同心圆法向量相互平行为约束条件确定工件真实位姿的方法。首先,通过圆位姿估计算法对摄像机采集的椭圆成像进行处理获取半径已知的空间圆的位姿二义解;其次,以标准差最小为评判依据,通过向量集搜索筛选法得出工件在空间中的真实位姿。经实验验证,工件的仰俯角及偏转角相对误差小于0.45°,计算简单,表明该方法简单可行,鲁棒性高。In order to estimate the pose of auto parts in automatic assembly technology,this paper proposes a method to determine the true pose of the workpiece,which based on the constraint condition that the normal vectors of different concentric circles in the same plane are parallel to each other.Firstly,the elliptical images collected by the camera were processed by the circle pose estimation algorithm to obtain the pose ambiguity solution of the spatial circle with known radius.Then,based on the minimum standard deviation,the real pose of the workpiece was obtained by the vector set searching and screening method.The experimental results show that the absolute error of the pitch and deflection angle is less than 0.45°,and the calculation is simple,which shows that the proposed method is simple and robust.

关 键 词：位姿估计 视觉检测 单目视觉 

分 类 号：TH165[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]