`

Timezone: »

 
Single Image Super-Resolution with Uncertainty Estimation for Lunar Satellite Images
Jose Delgado-Centeno · Paula Harder · Ben Moseley · Valentin Bickel · Siddha Ganju · Miguel Olivares · Freddie Kalaitzis

@
in Machine Learning and the Physical Sciences »

Recently, there has been a renewed interest in returning to the Moon, with many planned missions targeting the south pole. This region is of high scientific and commercial interest, mostly due to the presence of water-ice and other volatiles which could enable our sustainable presence on the Moon and beyond. In order to plan safe and effective crewed and robotic missions, access to high-resolution (<0.5 m) surface imagery is critical. However, the overwhelming majority (99.7%) of existing images over the south pole have spatial resolutions >1 m. In order to obtain better images, the only currently available way is to launch a new satellite mission to the Moon with better equipment to gather more precise data. In this work we develop an alternative that can be used directly on previously gathered data and therefore saving a lot of resources. It consist of a single image super-resolution (SR) approach based on generative adversarial networks that is able to super-resolve existing images from 1 m to 0.5 m resolution, unlocking a large catalogue of images (∼50,000) for a more accurate mission planning in the region of interest for the upcoming missions. We show that our enhanced images reveal previously unseen hazards such as small craters and boulders, allowing safer traverse planning. Our approach also includes uncertainty estimation, which allows mission planners to understand the reliability of the super-resolved images.

Author Information

Jose Delgado-Centeno (University of Luxembourg)
Paula Harder (Fraunhofer ITWM)
Ben Moseley (University of Oxford)
Valentin Bickel (Max Planck Institute for Solar System Research)
Siddha Ganju (Nvidia)
Miguel Olivares (Universidad de Luxemburgo)
Freddie Kalaitzis (University of Oxford)
Freddie Kalaitzis

Freddie is a part-time Senior Research Fellow, and Theme Lead of ML for Earth Observation and Remote Sensing, in the Oxford Applied and Theoretical Machine Learning lab (led by Yarin Gal) of Oxford University. He's also an ML & Project Lead at NASA's Frontier Development Lab (FDL), and the (part-time) ML Lead of Trillium Technologies , the R&D production company behind FDL. Since FDL US 2020, Freddie has been a ML & Project Lead for project Waters Of The United States (WOTUS), in partnership with the USGS, Planet, Maxar, Google Cloud and NVIDIA, towards the ultimate vision for mapping all flowing water on Earth , at near real-time, by fusing LiDAR sensors and daily very high resolution (VHR) satellite imagery. He started his journey with FDL 2019 as a mentor , helping teams super-resolve solar magnetograms and predict GPS disruptions induced by solar weather . Until April 2020, he was an Applied Research Scientist in the AI for Good lab (led by Julien Cornebise) of Element AI in London, focusing on applications of ML and statistics that enable NGOs and nonprofits. During this work, he led the Multi-Frame Super-Resolution research collaboration with Mila Montréal , which was awarded by ESA for topping the PROBA-V Super-Resolution challenge . He also co-authored the technical report written with Amnesty International, on the first large-scale study of online abuse against women on Twitter , whose front-page coverage in the Financial Times led to Twitter working with Amnesty to better protect the rights of vulnerable users online . Also with Amnesty, he has co-authored a technical blog-post on using VHR satellite imagery to detect evidence of genocide in rural areas of Darfur, Sudan. Prior to Element AI, he was a Senior Data Scientist in Digital Shadows, a Data Scientist in Microsoft's Xbox EMEA team, and a postdoc in Bayesian statistics (with Ricardo Silva) at the Statistical Science department of University College London . He has a PhD in Computer Science from the University of Sheffield (supervised by Neil Lawrence), and a MSc in Artificial Intelligence from the University of Edinburgh . His PhD research led to contributions in probability methods for dimensionality reduction of data and developed methods for gene-expression time-series to discover genetic factors of disease. Off work, he occasionally mentors teenagers of ages 12-18 for Teens in AI, and he is the founder of Well-Being in ML , the first official social event at NeurIPS advocating for healthy well-being and mental practice in academia and industry of ML .

More from the Same Authors