Advanced space robotic systems and spacecraft require three dimensional information to approach, inspect and service satellites, and assemble space structures. Such operations will require computer vision systems capable of computing pose (position and orientation) with high reliability and autonomy, and under difficult on-orbit illumination. Results of on-going research and development of computer vision techniques for space are described. The system uses 3D (Three Dimensional) surface models and sparse 3D data computed from stereo images to estimate the object pose. The approach is based on the ICP (Iterative Closest Point) algorithm and can deal with partial data loss due to illumination and occlusion. Different representations of models affect the accuracy, speed and reliability of the iterative alignment process. Experimental results and performance of pose estimation obtained for several representations are presented: models directly exported from CAD (Computer Aided Design) packages, regular triangular meshes, parametric and hybrid models.
Fast 3D pose estimation for on-orbit robotics
Schnelle Positions- und Orientierungsbestimmung aus Stereobildern für Roboter in Raumstationen
2000
7 Seiten, 3 Bilder, 1 Tabelle, 8 Quellen
Conference paper
English
Fast pose estimation with parameter-sensitive hashing
IEEE | 2003
|Fast Pose Estimation with Parameter-Sensitive Hashing
British Library Conference Proceedings | 2003
|High-accuracy real-time satellite pose estimation for in-orbit applications
Elsevier | 2025
|