光学 精密工程, 2014, 22 (5): 1165, 网络出版: 2014-06-03
大尺度钢板的三维测量和拼接
Three-dimensional measurement and registration for large scale plates
三维测量 三维拼接 计算机视觉 结构光 大尺度钢板 3D measurement 3D registration computer vision structured light large-scale plate
摘要
提出了基于投影的三维测量拼接方法,用于测量大尺度钢板表面的三维形状。首先,利用光学扫描仪、背景投影仪两种装置实现大尺度钢板的三维测量。其中,三维光学扫描仪负责测量大尺度钢板不同部分的三维数据,背景投影仪用于向被测钢板投射背景纹理;然后,利用基于随机抽样一致性算法(RANSAC)的拼接算法,将不同时刻测量的局部三维数据进行拼接,得到完整的钢板三维数据;最后,提出了一种拼接误差的评价方式来检验拼接精度。实验结果表明:所提方法的单次拼接精度为0.5 mm左右;测量一个7.5 m长的钢板,其累计拼接误差为2 mm左右。得到的结果基本满足船舶外板加工的精度要求,具有较高的实用价值。
Abstract
A 3D registration method based on the projection was proposed to align different 3D point clouds for measuring largescale bending steel plates.First, the 3D measurement of a largescale bending plate was performed by using a 3D optical scanner and a background projector, in which the 3D optical scanner was used to acquire the dense data of different partial views for the large bending plate and the background projector was used for projecting the background texture on the bending plate to align the point clouds measured at different time.Then,the point clouds were aligned by using the RANdom SAmple Consensus(RANSAC) based registration method and the 3D data measured from different time were registered.Finally, an evaluation method was proposed to verify the registration precision.The experimental results show that the precision of one registration is about 0.5 mm, and the accumulated registration error for measuring a plate with a length of 7.5 mm is about 2 mm.The effectiveness and performance evaluated by the experiments prove that the proposed method can satisfy the requirement of stability and accuracy for the industrial applications.
史金龙, 钱强, 庞林斌, 王直, 白素琴. 大尺度钢板的三维测量和拼接[J]. 光学 精密工程, 2014, 22(5): 1165. SHI Jin-long, QIAN Qiang, PANG Lin-bin, WANG Zhi, BAI Su-qin. Three-dimensional measurement and registration for large scale plates[J]. Optics and Precision Engineering, 2014, 22(5): 1165.