基于SFM的三维重建在点胶加工中的应用  

Application of 3D Reconstruction Based on SFM in Dispensing Processing

作　　者：洪国 张春雷 覃琴 HONG Guo;ZHANG Chunlei;QIN Qin(School of Mechanical Engineering,Sichuan University,Chengdu 610065,China;Chengdu Leetro Auto-Mation Co.,Ltd.,Chengdu 510107,China)

机构地区：[1]四川大学机械工程学院,成都610065 [2]成都乐创自动化技术有限公司,成都510107

出　　处：《组合机床与自动化加工技术》2024年第5期55-59,共5页Modular Machine Tool & Automatic Manufacturing Technique

摘　　要：为了实现点胶加工工件的三维模型重建,在分析了SFM(structure from motion)算法原理和五轴点胶机特点的基础上,提出了一种在五轴机床上使用单目相机实现工件三维模型重建的方法。首先,通过标定获得相机内外参,建立世界坐标系、机床坐标系与相机坐标系之间的关系;其次,从不同角度拍摄工件,与一般SFM不同的是拍摄时由机床机械坐标计算出相机位姿作为外参;最后,将图像序列和相机内外参作为输入,经过稀疏重建与稠密重建输出三维模型。实验结果表明,选用恰当的原始数据与参数可以高效地完成重建任务,重建成功率在90%以上,平均重投影误差为0.653 144 pixels。A method for reconstructing workpiece models using a monocular camera on a five-axis machine tool was proposed.The method analyzes the SFM(structure from motion)algorithm and the characteristics of a five-axis dispensing machine.Firstly,the camera′s internal and external parameters are obtained through calibration,and the relationship between the world coordinate system,machine coordinate system,and camera coordinate system is established;Then,the workpiece is photographed from different angles.Unlike general SFM,the camera pose is calculated from the mechanical coordinates of the machine tool as an external reference during the shooting process;Finally,using image sequence and camera parameters,a 3D model is obtained through sparse and dense reconstruction.The experimental results show that the reconstruction task can be efficiently completed by selecting appropriate raw data and parameters,the reconstruction success rate is more than 90%,and the average re projection error is 0.653144 pixels.

关 键 词：SFM 三维重建 五轴点胶机 机器视觉 

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]