航空学报 > 2023, Vol. 44 Issue (2): 326695-326695   doi: 10.7527/S1000-6893.2022.26695

基于旋转搜索的相机位姿估计和对应点匹配

王平1,2(), 付辉1, 徐贵力3   

  1. 1.兰州理工大学 电气工程与信息工程学院,兰州  730050
    2.甘肃省工业过程先进控制重点实验室,兰州  730050
    3.南京航空航天大学 自动化学院,南京  210016
  • 收稿日期:2021-11-24 修回日期:2021-12-27 接受日期:2022-05-18 出版日期:2023-01-25 发布日期:2022-06-08
  • 通讯作者: 王平 E-mail:pingwangsky@163.com
  • 基金资助:
    国家自然科学基金(62173170);甘肃省青年科技基金计划(20JR10RA186);流程工业综合自动化国家重点实验室联合开放基金(2021-KF-21-04);甘肃省工业过程先进控制重点实验室开放基金(2022KX02)

Camera pose estimation and corresponding points matching based on rotation search

Ping WANG1,2(), Hui FU1, Guili XU3   

  1. 1.College of Electrical and Information Engineering,Lanzhou University of Technology,Lanzhou  730050,China
    2.Key Laboratory of Gansu Advanced Control for Industrial Processes,Lanzhou  730050,China
    3.College of Automation Engineering,Nanjing University of Aeronautics and Astronautics,Nanjing  211106,China
  • Received:2021-11-24 Revised:2021-12-27 Accepted:2022-05-18 Online:2023-01-25 Published:2022-06-08
  • Contact: Ping WANG E-mail:pingwangsky@163.com
  • Supported by:
    National Natural Science Foundation of China(62173170);Gansu Province Science Foundation for Youths(20JR10RA186);State Key Laboratory of Synthetical Automation for Process Industries(2021-KF-21-04);Key Laboratory of Gansu Advanced Control for Industrial Processes(2022KX02)

摘要:

传统的同步位姿计算和对应点匹配(SPCD)算法采用两阶段交替迭代的方式计算相机位姿,求解精度较低。除此之外,传统SPCD算法依赖于局部搜索策略,无法保证找到全局最优解。为提升求解SPCD问题的精度和全局收敛能力,推导出了一种新的基于点和点约束的SPCD问题求解模型。基于新模型从角度距离出发构建优化目标函数,并利用分支定界搜索的方法寻找最优相机位姿和3D/2D匹配关系。实验证明:相比于传统SoftPOSIT算法,本文算法具有更高的求解精度和收敛率,并且可适用范围更广。

关键词: 计算机视觉, 视觉导航, 视觉测量, 位姿估计, PnP问题, SPCD问题

Abstract:

The traditional Simultaneous Pose and Correspondence Determination (SPCD) algorithm uses a two-stage alternate iteration strategy to calculate the camera pose, which leads to poor accuracy. Besides, the traditional SPCD algorithm relies on local search strategies, and have no guarantee of finding the globally optimal solution. To improve the accuracy and the ability of global convergence of solving the SPCD problem, a new solving model based on point-to-point constraints is derived for the SPCD problem in this paper. Based on this new model, an optimal objective function is constructed by using the angle distance, and the branch-and-bound method is used to search the optimal camera pose and the 3D/2D correspondences. Experiments show that the proposed method has higher accuracy and convergence rate compared to the traditional SoftPOSIT algorithm, and would have wide applications.

Key words: computer vision, vision-based navigation, vision measurement, pose estimation, PnP problem, Simultaneous Pose and Correspondence Determination (SPCD) problem

中图分类号: