Lunar Crater Identification via Deep Learning Ari Silburta,b,c,f, Mohamad Ali-Diba,d,f, Chenchong Zhub,d, Alan Jacksona,b,e, Diana Valenciaa,b, Yevgeni Kissinb, Daniel Tamayoa,d, Kristen Menoua,b aCentre for Planetary Sciences, Department of Physical & Environmental Sciences, University of Toronto Scarborough, Toronto, Ontario M1C 1A4, Canada bDepartment of Astronomy & Astrophysics, University of Toronto, Toronto, Ontario M5S 3H4, Canada cDepartment of Astronomy & Astrophysics, Penn State University, Eberly College of Science, State College, PA 16801, USA dCanadian Institute for Theoretical Astrophysics, 60 St. George St, University of Toronto, Toronto, ON M5S 3H8, Canada eSchool of Earth and Space Exploration, Arizona State University, 781 E Terrace Mall, Tempe, AZ 85287-6004, USA fThese authors contributed equally to this work. Keywords: Moon; Crater Detection; Automation; Deep Learning Corresponding authors: • Ari Silburt, Department of Astronomy & Astrophysics, Penn State Uni- versity, Eberly College of Science, State College, PA 16801, USA. e- mail:
[email protected] . • Mohamad Ali-Dib, Centre for Planetary Sciences, Department of Phys- ical & Environmental Sciences, University of Toronto Scarborough, Toronto, Ontario M1C 1A4, Canada. e-mail:
[email protected] . arXiv:1803.02192v3 [astro-ph.EP] 12 Nov 2018 Preprint submitted to Elsevier November 13, 2018 Abstract Crater counting on the Moon and other bodies is crucial to constrain the dynamical history of the Solar System. This has traditionally been done by visual inspection of images, thus limiting the scope, efficiency, and/or accuracy of retrieval. In this paper we demonstrate the viability of using convolutional neural networks (CNNs) to determine the positions and sizes of craters from Lunar digital elevation maps (DEMs).