Extraction of Common Regions of Interest from Orbital and Rover (Ground) acquired Imagery

Accurate global localization of space exploratory rovers is of high importance to the future extraterrestrial missions. In particular, the Mars Sample Return mission (MSR) needs precise localization techniques for the Sample Fetching Rover (SFR) to collect a previously drilled cache of soil. Moreover, exact positioning is sine qua non for a successful rendezvous of the SFR with a Mars Ascent Vehicle (MAV) that will launch the load into Martian orbit. The utilization of prior information to refine the localization has been the main approach in mobile robotics during the past decades. In extraterrestrial scenery, orbital images constitute an excellent source of such prior information. With long term aim to provide accurate global rover localization, as a first step, the proposed paper assesses the ability to automatically detect common regions of interest from both rover (ground) and orbital images. The approach comprises the separate extraction of salient regions, employing distinct techniques for the orbital and ground images.