Efficient Feature Matching and Mapping for Terrain Relative Navigation Using Hypothesis Gating

Terrain relative navigation systems commonly leverage optical imaging cameras for visual measurements. These measurements typically take the form of matching smaller a priori template images to a spacecraft cameras field-of-view, which aid the onboard localization process in estimating the position of the spacecraft. It would be more beneficial to gain this localization knowledge through processing the stream of images directly without any a priori knowledge, by employing traditional image processing techniques such as feature extraction and matching. This process works well for navigation in terrestrial settings (like Visual SLAM), but can be challenging in space where poor lighting and limited visual variety are observed, effecting the quality of the matches obtained. Further, as spacecraft compute power typically lags behind the latest state of the art, computational intensity imposed by the feature matching process is typically too great to be ran onboard spacecraft in real-time. This paper tackles the inaccuracies and inefficiencies of standard image feature matching processes in space environments, by leveraging traditional onboard navigation filter information to drastically reduce the number of matching candidates. Estimated feature location is used to form statistical prediction gates around a given feature, for which all points lying inside are treated as inliers and fed to the matching process. Using a simulated trajectory around a high-fidelity 3D asteroid model and a single monocular camera, we demonstrate an overall reduction of around 87% in average matching time for three popular feature description techniques. A substantial increase in the quality of matches obtained is also shown, giving utility towards purely monocular terrain relative navigation.