基于机器视觉的印刷机轴套工装位姿测量

Machine Vision Based Pose Estimation of Printing Machine Shaft Sleeve Tooling

摘要: 针对印刷机轴承套筒装配过程中无约束运动阶段位姿测量问题，以Stewart平台为装配操作平台，通过提取双目相机下的图像椭圆特征，设计出一种基于机器视觉的印刷机轴套工装位姿测量系统。首先以轴套工装外径与相机成像平面之间投影形成的椭圆特征为识别目标，利用投影几何关系估算轴套工装位姿得到二义性解，用双目视觉定位技术剔除伪解，最后根据位姿变换得到工装位姿并且反解出六组支杆杆长。实验结果表明，杆长最大误差率不超过1.84%，能为无约束运动阶段轴孔装配提供位姿信息和杆长控制，进而实现印刷机轴套装配中无约束运动阶段的轴孔装配。

Abstract: Aiming at the pose measurement problem during the unconstrained motion stage of the printing machine bearing sleeve assembly process, using the Stewart platform as the assembly operation platform, a machine vision based printing machine shaft sleeve tooling pose measurement system is designed by extracting image elliptical features under binocular cameras. First and for most, estimate the pose of this sleeve fixture using geometric relationships. There will be two pose solutions in this step. Then, use binocular vision positioning technology to eliminate erroneous pose solutions. Finally, based on the pose transformation, the fixture pose is obtained and six sets of support rod lengths are solved in reverse. The experimental results show that the maximum error rate of the rod length does not exceed 1.84%, which can provide pose information and rod length control for the assembly of the shaft hole in the unconstrained motion stage. Thus, it achieves the assembly of the shaft hole in the unconstrained motion stage of the printing machine shaft assembly.