Using One-Dimensional Signal Features for Image-Based Lunar Skyline Recognition

As lunar surface exploration becomes a higher priority there is increased need for methods of lunar surface navigation. Autonomous navigation (i.e., without reliance on human operators or other assets) will become essential for safe and robust surface operations without overburdening Surface Communications and Navigation (SCAN) resources. A recent development in this field is the utilization of the visible skyline for localization, either by comparison to pre-rendered skylines or to skylines rendered in real-time. This requires a fast and accurate way to compare two skylines (rendered and observed) and to rank skyline similarity.

This work develops a novel one-dimensional (1D) signal feature. This is the lower-dimensional analog of familiar 2D image features, often used in image processing and computer vision. This signal feature is then used in the context of finding common points between disparate sets of skyline points, which are expressed as sets of corresponding azimuth and elevation angle pairs. This feature can be used to rank the similarity between skylines rendered at different locations on the lunar surface. It is shown that this technique is less computationally expensive and has higher accuracy when compared to other skyline similarity metrics. Finally, these 1D features are demonstrated in the context of a full surface navigation simulation using Apollo 17 images and data.