测绘学报 ›› 2026, Vol. 55 ›› Issue (3): 465-476.doi: 10.11947/j.AGCS.2026.20250287

• 大地测量学与导航 • 上一篇    下一篇

基于先验位姿与运动编排的相机惯导外参标定方法

周瑞1(), 朱锋1(), 张小红2   

  1. 1.武汉大学测绘学院,湖北 武汉 430079
    2.武汉大学中国南极测绘研究中心,湖北 武汉 430079
  • 收稿日期:2025-07-14 修回日期:2026-03-18 出版日期:2026-04-16 发布日期:2026-04-16
  • 通讯作者: 朱锋 E-mail:ruichou@whu.edu.cn;fzhu@whu.edu.cn
  • 作者简介:周瑞(2000—),男,硕士生,主要研究方向为多源融合导航定位。E-mail:ruichou@whu.edu.cn
  • 基金资助:
    国家杰出青年科学基金(42425003);国家自然科学基金(42388102; 42374031)

Camera-IMU extrinsic calibration based on prior poses and motion planning

Rui ZHOU1(), Feng ZHU1(), Xiaohong ZHANG2   

  1. 1.School of Geodesy and Geomatics, Wuhan University, Wuhan 430079, China
    2.Chinese Antarctic Center of Surveying and Mapping, Wuhan University, Wuhan 430079, China
  • Received:2025-07-14 Revised:2026-03-18 Online:2026-04-16 Published:2026-04-16
  • Contact: Feng ZHU E-mail:ruichou@whu.edu.cn;fzhu@whu.edu.cn
  • About author:ZHOU Rui (2000—), male, postgraduate, majors in multi-sensor fusion navigation and positioning. E-mail: ruichou@whu.edu.cn
  • Supported by:
    The National Science Fund for Distinguished Young Scholars of China(42425003);The National Natural Science Foundation of China(42388102; 42374031)

摘要:

多传感器融合利用异构传感器观测的互补性实现高精度导航定位,而精确的传感器外参是实现可靠融合的前提。本文针对现有相机惯导外参标定方法中对标定板依赖较强、标定结果易受数据质量影响、数据采集过程较为复杂等问题,提出一种基于先验位姿与运动编排的无标靶相机惯导外参标定方法:利用GNSS/SINS后处理平滑结果作为先验位姿,构建重投影误差方程,基于高斯牛顿优化框架实现外参的精确估计;通过运动编排优化数据采集轨迹,利用运动转台充分激励惯导,提高影像重叠度,保证标定数据的重复性和可靠性;提出一套完整的初始化与外参精优化流程,加速优化过程的收敛,实现外参的最优估计。仿真结果表明,该方法可实现优于0.05°的旋转外参估计精度和优于1 cm的平移外参估计精度;实测试验验证了该标定方法及流程的可行性与标定效果。

关键词: 组合导航, 传感器标定, 惯性导航系统, 多源融合

Abstract:

Multi-sensor fusion leverages the complementarity of heterogeneous data to achieve high-precision navigation and positioning, where accurate sensor extrinsic calibration serves as a fundamental prerequisite. To address the limitations of traditional camera-IMU calibration methods—such as strong reliance on calibration targets, sensitivity of calibration performance to data quality, and cumbersome data acquisition—this paper proposes a targetless calibration algorithm based on prior poses and motion planning. The method utilizes GNSS/SINS post-processed smoothed trajectories as prior poses to construct a reprojection error model, and employs a Gauss-Newton optimization framework to estimate the extrinsic parameters with high accuracy. A motion planning strategy is developed to optimize the data acquisition trajectory, using a motion turntable to sufficiently excite the IMU and enhance image overlap, thereby ensuring the repeatability and reliability of the calibration data. Furthermore, a complete initialization and extrinsic refinement pipeline is introduced to accelerate convergence and achieve optimal calibration results. Simulation results demonstrate that the proposed method achieves an extrinsic calibration accuracy better than 0.05° in orientation and 1 cm in translation. Field experiments further validate the feasibility and effectiveness of the proposed calibration approach and workflow.

Key words: integrated navigation, sensor calibration, inertial navigation system, multi-sensor fusion

中图分类号: