This work proposes a novel visual pose tracking method for on-orbit uncooperative targets by making use of fused data from 3D laser scanner and optical camera. In particular, the target point cloud from the laser scanner is precisely segmented by K-means cluster methods, and projected on the pixel grid to extract accurate depth information. The obtained depth information is fused with the target feature point of the optical images, which then is used to estimate the relative pose for tracking tasks. Different with the RGB-D camera or other sensor fusion method, the proposed tracking method can maintain a robust and accurate tracking performance even in presence of large background interference in the images. The proposed visual tracking method was comprehensively tested with two different datasets, the synthetic images without background interference and the standard satellite dataset with large background interference. With comparisons with other sensor fusion methods based on images and 3D point cloud, the proposed fusion method show more robust and accurate performance.
Accurate Pose Tracking for Uncooperative Targets via Data Fusion of Laser Scanner and Optical Camera
J Astronaut Sci
The Journal of the Astronautical Sciences ; 69 , 5 ; 1375-1393
2022-10-01
19 pages
Aufsatz (Zeitschrift)
Elektronische Ressource
Englisch