Author(s): Georgios Georgakis, Jet Propulsion Laboratory, California Institute of Technology (First Author, Presenting Author)
The need to extend operational capabilities of autonomous planetary explorers and the increase of complexity and volume of scientifically valuable mission data requires us to explore novel deep learning solutions to problems that typically have been driven by traditional image processing algorithms. In our work, we explore two such use cases. First, we formulate the photometric calibration of JunoCam as an image-to-image translation problem and evaluate the validity of data-driven calibrations. Second, we enable long-range aerial traverses on Mars via deep image registration of onboard Ingenuity images to orbital maps and explore the robustness of deep learning under large resolution differences.
