多镜头组合型全景相机的位姿估计算法  

Pose Estimation Algorithm For Multi-lens Combined Panoramic Camera

作　　者：杨荣淇 张文涛 许金城 尹双双 YANG Rongqi;ZHANG Wentao;XU Jincheng;YIN Shuangshuang

机构地区：[1]上海船舶设计研究院 [2]武汉大学测绘学院

出　　处：《计量与测试技术》2023年第3期36-40,共5页Metrology & Measurement Technique

摘　　要：本文针对多镜头组合型全景相机在定向中面临的拼接误差、匹配误差和前方交会误差控制等难题,通过多镜头组合型相机之间刚体不变的相对位姿约束,提出Eye-to-Eye相机标定方法恢复统一的场景坐标系,采用多镜头SfM算法的后方交会,实现了更多的位姿估计。实验结果表明:该方法能成功实现所有影像定向,并能较好地恢复三维结构。In this paper,aiming at the problems such as splice error,matching error and front alignment error control in orientation of multi-lens combined panoramic cameras,the Eye-to-Eye camera calibration method is proposed to restore a unified scene coordinate system through the rigid body invariant relative pose constraint between multi-lens combined panoramic cameras,and the multi-lens SfM algorithm is used for rear alignment to realize more pose estimation.Experimental results show that this method can achieve all image orientation successfully and restore 3D structure well.

关 键 词：多镜头组合型 全景相机 SfM算法 位姿估计 

分 类 号：TB852.1[一般工业技术—摄影技术]