长短焦双目三维成像算法研究与精度分析  

Research and precision analysis on binocular 3D imaging algorithm with long and short focal length

作　　者：李仁杰 阳波[2] 肖宏峰 陈灵 邹雅萍 唐文胜 LI Renjie;YANG Bo;XIAO Hongfeng;CHEN Ling;ZOU Yaping;TANG Wensheng(School of Information Science&Engineering,Hunan Normal University,Changsha 410008,China;School of Engineering&Design,Hunan Normal University,Changsha 410008,China)

机构地区：[1]湖南师范大学信息科学与工程学院,湖南长沙410008 [2]湖南师范大学工程与设计学院,湖南长沙410008

出　　处：《传感器与微系统》2024年第9期42-46,共5页Transducer and Microsystem Technologies

基　　金：国家自然科学基金资助项目(62072175)。

摘　　要：针对复杂三维装配过程中提高三维位置精度会导致视野变窄造成三维装配定位失败的问题,提出一种基于自适应缩放因子估计的长短焦距双目三维成像算法。该算法先根据长短焦相机获取2幅图像的特征匹配结果估计特征尺度缩放因子并且实现特征尺度的归一化,再通过半全局立体匹配算法得到双目视差,最后计算出目标的三维点云。长短焦距成像和同焦距成像精度与视野大小的对比实验结果表明,该算法在感兴趣区域内与同焦距双目系统计算视差的最小错误估计比例小于2.6%,系统观测视野扩大32%。Aiming at the problem that the improvement of 3D position precision in the process of complex 3D assembly will lead to the narrowing of visual field and results in failure of 3D assembly positioning,a long and short focal length binocular 3D imaging algorithm based on adaptive scaling factor estimation is proposed.Contrast experiment of precision of long and short focal length imaging and same focal imaging and field of view(FOV)size is carried out.The algorithm firstly estimates the feature scale factor and realizes the normalization of the feature scale according to the feature matching results of the two images obtained by the long and short focal camera.Then the binocular parallax is obtained by the semi-global stereo matching algorithm.Finally,the 3D point cloud of the target is calculated.Contrast experiment of precision of long and short focal length imaging and same focal imaging and FOV size is carried out.The experimental results show that the minimum error estimation ratio of parallax calculation between the algorithm and the binocular system with the same focal length in the region of interest is less than 2.6%,the system observation field is enlarged by 32%.

关 键 词：双目视觉 自动装配 立体匹配 缩放因子 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]