未知主点条件的相机焦距自标定方法  

Focal length self-calibration with unknown principal point from two views

作　　者：唐荣富 邓宝松 赵竹新 Tang Rongfu;Deng Baosong;Zhao Zhuxin(National Innovation Institute of Defense Technology,Academy of Military Sciences,Beijing 100071,China;Beijing Remote Sensing Institute,Beijing 100192,China)

机构地区：[1]军事科学院国防科技创新研究院,北京100071 [2]北京遥感研究所,北京100192

出　　处：《计算机应用研究》2020年第4期1226-1229,共4页Application Research of Computers

基　　金：军队应用基础研究项目。

摘　　要：对于双目相机焦距自标定,当前研究方法均假定相机主点位置已知,而相机主点在实际中却通常是未知的,提出一种主点和场景均未知的条件下相机焦距自标定的新方法,通过严格的数学分析表明相机的坐标系尺度对焦距标定具有重要影响,并在坐标系尺度影响的定量分析的基础上提出了一种选择合适的坐标系尺度算法。模拟和实际实验都验证了坐标系尺度的重要性,且该方法较传统方法可获得更好的标定精度。It is known that focal length can be recovered from two views when other intrinsic parameters are fixed.This paper presented a more practical work to calibrate focal length from two views with unknown principal point(neither structure nor motion was known a priori).The mathematical analyses showed that the scale of camera coordinates played a crucial role in such calibration circumstance.It proposed a heuristic algorithm to select appropriate scale factor for calibrating focal length.Both simulated and practical tests confirm the significance of scale factor and the good performance of the proposed algorithm,which obtains more accurate focal length calibration than those with no account of scale factor.

关 键 词：计算机视觉 相机自标定 焦距 坐标系尺度 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]