A ROBUST IMAGE PROCESSING PIPELINE FOR PLANETS LINE-OF-SIGHT EXTRACTION FOR DEEP-SPACE AUTONOMOUS CUBESATS NAVIGATION

The adoption of CubeSats marks the beginning of new momentum in interplanetary exploration, owing to their low cost as compared to standard missions. Yet, the proliferation of deep-space CubeSats will cause the saturation of the ground networks, hindering the traditional navigation through ground-based radiometric tracking. The solution to this issue resides in the exploitation of autonomous miniaturized probes. Autonomous vision-based navigation (VBN) and robust Image Processing (IP) are crucial to overcome these limitations. This is because of the capability of such algorithms to provide near real-time information to navigation filters on-board. For these reasons the use of passive cameras, in combination with IP algorithms, provides compelling navigation performances with light and cost-effective hardware. It is thus necessary to develop efficient, robust, and fast IP algorithms that must be strongly integrated within navigation filters. In the context of deep space navigation, CubeSats can perform celestial triangulation by exploiting visible planets as beacons. Given these considerations, this paper proposes an innovative and robust planets line-of-sight extraction procedure applicable to interplanetary autonomous CubeSats. In particular, the work focuses on the description of a new methodology adopted for star pattern and planets identification. Statistical results show that a 3σ accuracy of 30 arcsec is obtained for the planet line-of-sight (LoS) extraction with a failure lower than 1% when the spacecraft position is known with a 3σ accuracy of 105 km.