无编码点的工业摄影测量技术的研究及实现  

Research and Implementation of Industrial Photogrammetry Without Coded Points

作　　者：严俊 叶南[1] 李廷成 祝鸿宇 Yan Jun;Ye Nan;Li Tingcheng;Zhu Hongyu(College of Mechanical and Electrical Engineering,Nanjing University of Aeronautics and Astronautics,Nanjing,Jiangsu210016,China)

机构地区：[1]南京航空航天大学机电学院

出　　处：《光学学报》2019年第10期219-228,共10页Acta Optica Sinica

基　　金：国家自然科学基金(51605222);江苏省基础研究计划(BK20160799);上海航天科技创新基金资助项目

摘　　要：现有的工业摄影测量系统通常需要在被测物表面布设一定数量的编码点,但是许多工业产品表面并不具备布设编码点的条件。为此,提出一种无需布设编码点的工业摄影测量的实现方法。该方法只需利用一台向被测物投射散斑纹理的投射装置、一把恢复尺度的标尺,以及一台能对被测物体多角度拍摄的相机。为了完成高精度的多视图几何框架的求解,采用一种"由粗到精"的两步重建策略,并提出一种带旋转补偿的数字图像相关(RFDIC)方法来解决大旋转角度图像间的高精度匹配问题。实验结果表明,RFDIC方法对标尺长度测量的误差小于0.01 mm/m,与最新商用的三维测量系统点云重建算法相比,二者的误差约为0.055 mm,满足工业摄影测量的精度要求。Placing a certain number of coded points is generally required in the current industrial photogrammetry system. However, many industrial products are not suitable for the arrangement of coded points. This study proposes a novel method of industrial photogrammetry without using coded points. Our method only requires a projection device to project the speckle texture, a scale bar to recover scale, and a multi-angle camera to capture various images of the measured object. A "coarse-to-fine" two-step reconstruction strategy is devised to solve the multi-view geometry with high precision. Moreover, a rotation-free digital image correlation(RFDIC) method is proposed for high-accuracy point-matching between images with large rotational angles. The experimental results verify that the error in measuring the length of scale bar by the RFDIC method is below 0.01 mm/m and the error of the RFDIC method compared with that of the latest commercial point cloud construction method for three dimensional measurement system can reach approximately 0.055 mm, which satisfies the precision requirements of industrial photogrammetry.

关 键 词：机器视觉 无编码点摄影测量 投影散斑 数字图像相关 多视图几何 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]