Monocular SLAM with Point and Line Features Applied to Spacecraft Relative Navigation

• Published in: International journal of aeronautical and space sciences Vol. 26; no. 3; pp. 1258 - 1278
• Main Authors: Pan, Ruitao, Wang, Chenxi, Zhai, Zhi, Liu, Jinxin, Pan, Tianhang, Chen, Xuefeng
• Format: Journal Article
• Language: English
• Published: Seoul The Korean Society for Aeronautical & Space Sciences (KSAS) 01.04.2025; Springer Nature B.V; 한국항공우주학회
• Subjects: Aerospace environments; Aerospace Technology and Astronautics; Algorithms; Cameras; Engineering; Feature extraction; Fluid- and Aerodynamics; Hypotheses; Kinematics; Methods; Navigation; Original Paper; Pose estimation; Real time; Satellites; Simultaneous localization and mapping; Space rendezvous; Space telescopes; Spacecraft; State estimation; 항공우주공학; Visual SLAM; Uncooperative and unknown targets; Line features; Relative navigation; Pose estimation
• ISSN: 2093-274X; 2093-2480
• DOI: 10.1007/s42405-024-00817-2

Real-time estimation of the target’s pose is crucial for spacecraft relative navigation. If the target is uncooperative and unknown, i.e., with no prior information, the simultaneous localization and mapping (SLAM) technique is utilized to estimate both the target’s pose and 3D shape. Several point-feature-based methods, such as ORB-SLAM, have recently been tested for spacecraft rendezvous. However, point features perform poorly in weak-textured targets and illumination changes, commonly appearing in space environments. This paper presents a monocular SLAM system using point and line features for spacecraft relative navigation. The strengths of different features are fully explored. Specifically, the line feature extraction and matching algorithms are improved, and Plücker coordinates for line representation are used to solve the endpoint inconsistency problem. Moreover, the smoothing approach is utilized for better state estimation while the real-time performance is guaranteed. Compared to the ORB-SLAM, our method is more robust and accurate in complex space environments. Experiments on a challenging dataset demonstrate that adding line features can improve the system’s robustness and pose estimation accuracy by 63.6% with a better 3D shape reconstruction. The algorithm runs at 17.83 Hz, satisfying the real-time requirement.