Sovremennye problemy distantsionnogo zondirovaniya Zemli iz kosmosa, 2020, Vol. 17, No. 7, pp. 58-66
Specific features of pattern recognition problems in relation to the tasks of relative navigation during spacecraft docking
V.A. Grishin
1, 2 , B.S. Zhukov
1 1 Space Research Institute RAS, Moscow, Russia
2 Bauman Moscow State Technical University, Moscow, Russia
Accepted: 17.11.2020
DOI: 10.21046/2070-7401-2020-17-7-58-66
Relative navigation during docking is carried out in space environment, which leads to a number of specific features that must be taken into account when selecting image processing and recognition methods. In space, there is no atmosphere that produces diffuse illumination of objects due to scattered radiation. Therefore, a typical situation in space is the presence of very deep shadows in spacecraft images, which are created by structural elements of the same spacecraft. The dynamic range of real cameras is usually completely insufficient to produce reasonable images in the shadows. Part of the spacecraft’s surface is covered with screen-vacuum thermal insulation, which can give bright blicks in the image. Also bright glare can be produced by the surface of solar panels. The spacecraft can be observed against the background both of the black sky and of the Earth. In the first case, the contrast is very high compared to the background, and in the second case it can be relatively low. Accordingly, different algorithms should be used to distinguish the spacecraft against the background of the sky and the Earth. An arbitrary angle of observation of a passive spacecraft during uncooperative docking necessitates the use of a two-stage recognition procedure. At the first stage, a rough definition of the spacecraft aspect angles and distance is made. The task of the second step is to accurately determine the spacecraft relative pose (position and orientation). A rough definition of the angle and distance can be carried out by invariant moments of binarized images, or by their silhouettes. Structural-linguistic methods can be used both to roughly determine the angle and distance and to accurately determine the spacecraft relative position and orientation.
Keywords: cooperative and non-cooperative docking of spacecraft, pattern recognition, measurement of relative position and orientation
Full textReferences:
- Pirat C., Ankersen F., Walker R., Gass V., Vision Based Navigation for Autonomous Cooperative Docking of CubeSats, Acta Astronautica, 2018, Vol. 146, pp. 418–434.
- Satellite Servicing, Solutions for Commercial Space and Other Applications, National Aeronautics and Space Administration, 2018, 8 p., available at: https://nexis.gsfc.nasa.gov/documents/SSPD_Pamphlet4_FINAL.pdf (accessed 21.10.2020).
- Telaar J., Ahrns I., Estable S., Rackl W., De Stefano M., Lampariello R., Santos N., Serra P., Canetri M., Ankersen F., Gil-Fernandez J., GNC Architecture for the e.Deorbit Mission, Proc. 7th European Conf. Aeronautics and Space Sciences (EUCASS 2017), Milan, Italy, 2017, 15 p., DOI: 10.13009/EUCASS2017-317.
- Toshev A., Makadia A., Daniilidis K., Shape-Based Object Recognition in Videos Using 3D Synthetic Object Models, Proc. 2009 IEEE Conf. Computer Vision and Pattern Recognition, Miami, Florida, 2009, pp. 288–295.
- Underwood C., Pellegrino S., Lappas V., Bridges C., Baker J., Using CubeSat/Micro-Satellite Technology to Demonstrate the Autonomous Assembly of a Reconfigurable Space Telescope (AAReST), Acta Astronautica, 2015, Vol. 114, pp. 112–122.