Satellite communications innovator and space gateway Goonhilly Earth Station has been assisting the University of Hertfordshire with the use of Goonhilly's deep learning platform and incubator hub to radically accelerate time to market for a new venture that will provide regular satellite mapping of the UK free from cloud cover for the first time.
The University of Hertfordshire team uses satellite radar imaging to generate detailed images of the Earth’s surface in the more familiar visible and infrared bands. Since radar can pass through clouds, this allows the team to map the entire UK on a regular basis, uninterrupted by cloud cover. The venture originated from work funded by the Science and Technology Facilities Council (STFC) and incorporates techniques developed to analyse the light from distant galaxies. The project is led by astrophysicist Professor James Geach and his Ph.D. student Mike Smith.
With the UK land mass often shrouded in cloud, this breakthrough enables a whole new sphere of business opportunities exploiting Earth Observation data. Target customers will include insurance firms, commodity traders, supermarkets and the agricultural industry.
Using Goonhilly’s NVIDIA GPU supercomputer, the University of Hertfordshire team trained its algorithm – called “ClearSky” – on a huge data set comprising tens of thousands of images of the UK taken by ESA’s Sentinel satellites, part of the Copernicus Earth Observation program. The team is already working with agritech company Agrimetrics to develop a pipeline for monitoring the health and growth rate of each of the UK’s 2.8 million fields on a weekly basis.
With every image pixel representing a 10×10 metre square of the UK, other applications are already under development. These include: predicting environmental threats such as flooding and wildfires; monitoring coastal erosion; and tracking the impact of climate change on crop growth patterns. Started as a research project, the team is now looking for further investment to spin out a commercial entity, “DeepEO”, later this year.
Building on recent momentum, Goonhilly is expanding its deep learning platform in a bold bid to become the UK’s premier incubator hub with deep learning specialties including a focus on Earth Observation data. It is now building a testbed using a variety of platforms, offering customers greater choice. Goonhilly’s High Performance Computing green data centre complements the deep learning platform, providing edge computing for near real-time storage and analysis of live satellite data streams.
Using Goonhilly’s platform, DeepEO plans to create a continuously updated ‘living’ Earth Observation database combining land, ocean and atmospheric data. The goal is to apply more innovative analysis techniques to derive valuable intelligence from the data and deliver this to end users. Customers will be able to accelerate and improve their decision making — for example, predicting in advance when a river will burst its banks — and take appropriate pre-emptive action.
Professor Geach stated that Goonhilly’s deep learning platform has allowed them to massively accelerate time to market. The platform’s phenomenal processing speed has made it possible for them to significantly scale up their models and increase the scope of their analysis. It’s rewarding to see how techniques developed for astrophysics can be applied to Earth Observation data to deliver real-world impact, and they’re excited by the range of possible applications.
Chris Roberts, Head of Data Centre and Cloud at Goonhilly added that their wraparound service nurtures start-ups like DeepEO with the resources they need on their journey from a research project to commercial growth and profitability. And their data center’s green credentials are an increasingly important factor, ensuring that valuable environmental efforts to mitigate the effects of climate change aren’t themselves creating volumes of CO2.
