盘盖类工件自动化装配视觉引导方法研究  

Research on Visual Guidance Method for Automated Assembly of the Disc-and-Cover-Like Workpieces

作　　者：陆雨薇 尹利钊 罗捷 施显柳 李远智 LU Yu-wei;YIN Li-zhao;LUO Jie;SHI Xian-liu;LI Yuan-zhi(Guangxi University of Science and Technology,Guangxi Key Laboratory of Automobile Components and Vehicle Technology,Liuzhou 545006;SAIC GM Wuling Automobile Co.,Ltd.,Guangxi Laboratory of New Energy Automobile,Liuzhou 545616;Liuzhou Huxin Automobile Technology Co.,Ltd.,Liuzhou 545616;School of Mechanical Engineering,Shanghai Jiao Tong University,Shanghai 200240)

机构地区：[1]广西科技大学广西汽车零部件与整车技术重点实验室,广西柳州545006 [2]上汽通用五菱股份有限公司广西新能源汽车实验室,广西柳州545007 [3]柳州沪信汽车科技有限公司,广西柳州545007 [4]上海交通大学机械与动力工程学院,上海200240

出　　处：《制造业自动化》2025年第1期20-28,共9页Manufacturing Automation

基　　金：广西科技计划《中国-东盟绿色车辆研究院搭建与能力建设》项目(桂科AA24206060);博士后专项基金(277291)。

摘　　要：为解决自动化装配过程中工件定位难,安装精度差等问题,设计了一种对点云质量适应性强的高精度、高效率的盘盖类工件自动化装配视觉引导系统。首先使用点云包围盒提取点云数据ROI,并对其进行去噪及降采样的预处理步骤。接着提出了一种使用有限元网格中心点制作点云模板的方法,并对PCA算法进行了主轴及质心矫正的改进,提高了算法对点云质量的容忍度,实现了点云数据快速准确配准。最后对点云模板制作方法的有效性、改进PCA算法的配准效果、算法对点云质量的容忍度、算法对不同零件的鲁棒性及系统整体精度进行了实验验证。系统最终的平移误差均值为0.80 mm,RMSE小于0.089 mm;旋转误差均值为1.24 mm,RMSE小于0.116 mm,完全满足了自动化装配的要求。实验结果表明,所提出的方法在点云质量容忍度、配准的精度及效率方面具有优越的综合性能。To solve the problems of difficult workpiece positioning and poor installation accuracy in the automated assembly process,a high-precision and high-efficiency visual guidance system was designed for automatic assembly of the plate-like parts with strong adaptability to point cloud quality.Firstly,the bounding boxes of point clouds were used to extract point cloud data ROI,and the preprocessing steps for denoising and down-sampling were performed.Subsequently,a method of using finite element mesh center points to create point cloud templates was proposed,and the PCA algorithm was enhanced with spindle and centroid correction to improve the algorithm’s tolerance for point cloud quality and achieve fast and accurate registration of the point cloud data.Finally,the effectiveness of the point cloud template creation method,the improved registration effect of the PCA algorithm,the algorithm’s tolerance for variations in point cloud quality,its robustness to different parts and the overall accuracy of the system were experimentally verified.Eventually,the average translation error of the system is 0.80mm,with the RMSE less than 0.089mm,and the average rotation error is 1.24mm,with the RMSE less than 0.116mm, indicating that the requirements of automated assembly were fully met. The experimental results show that the proposed method has superior comprehensive performance in terms of point cloud quality tolerance, regis‐tration accuracy and efficiency.

关 键 词：视觉引导 结构光相机 点云配准 

分 类 号：TP274.5[自动化与计算机技术—检测技术与自动化装置]