摘要
针对具有平面特征的工件的位姿识别,提出了一种三维点云两步校准法。首先,对三维点云进行多次随机抽样,每次抽取3个点并计算其所构成平面的单位法向量;经多次抽样后,得到一个平面单位法向量集合,采用密度聚类算法提取主平面的单位法向量,并根据主平面单位法向量与目标向量的轴角变换关系实现点云的倾斜度校正。然后,先将点云投影至主平面,再将各点的高度映射为主平面内的灰度,得到点云灰度图;再利用图像模板匹配算法获取点云在主平面内投影的平移量和旋转角,经平移和旋转后实现点云位姿的最终校准。将所提出的三维点云两步校准法应用于高压输电线塔塔座自动焊接系统,以识别塔座的位姿信息,并与基于采样一致性法识别的塔座位姿信息进行对比。结果表明,相比于采样一致性法,所提出的两步校准法既提高了工件位姿的识别精度,又缩短了约76%的耗时。研究结果可为具有平面特征的工件的位姿快速校准及识别提供参考。三维点云两步校准法在先进制造业中具有广阔的应用前景。
Aiming at the pose identification of workpieces with planar geometric features,a two-step alignment method for 3-D point cloud was proposed.Firstly,the 3-D point cloud was randomly sampled multiple times,while three points were sampled each time to calculate the unit normal vector of the plane formed by them.A set of unit normal vectors for planes was obtained after multiple sampling,and the unit normal vector of principal plane was extracted by the density clustering algorithm.The inclination correction of point cloud was achieved according to the axis angle transformation relationship between the unit normal vector of principal plane and the target vector.Then,the point cloud was projected to the principal plane,and the height of each point was mapped into the gray of principal plane to obtain the point cloud greyscale image.The image template matching algorithm was used to get the translation amount and rotation angle of the point cloud projection in the principal plane,and the final alignment of the point cloud pose was realized after translation and rotation.The proposed two-step alignment method for 3-D point cloud was applied to the automatic welding system for tower base of high-voltage transmission line to identify the pose of tower base,and a comparison was performed with the pose of tower base identified based on the sample consensus method.The experimental results proved that compared with sample consensus method,the proposed twostep alignment method not only improved the identification accuracy of workpiece pose,but also shortened the time consumption by about 76%.The research results provide reference for the rapid alignment and identification of poses of workpieces with planar features,and the two-step alignment method for 3-D point cloud has a wide application prospect in the advanced manufacturing.
作者
汪威
张开颜
刘亚川
黄玉春
WANG Wei;ZHANG Kai-yan;LIU Ya-chuan;HUANG Yu-chun(School of Mechanical Engineering,Hubei University of Technology,Wuhan 430068,China;School of Remote Sensing and Information Engineering,Wuhan University,Wuhan 430068,China)
出处
《工程设计学报》
CSCD
北大核心
2020年第5期560-567,共8页
Chinese Journal of Engineering Design
基金
国家自然科学基金资助项目(61976083)。
关键词
点云校准
倾斜度校正
点云灰度图
图像匹配
point cloud alignment
inclination correction
point cloud grayscale image
image matching
