Earth observation (EO) data is continuously being acquired through remote-sensing satellites to monitor biological, chemical and physical processes which characterise our planet, as well as the impact of human activities and climate change.
Through the Sentinel satellite missions, the Copernicus Programme acts as "Europe's eyes on Earth" to gather such data, on top of which a number of pipelines can be built for several monitoring tasks, from land cover mapping to precision agriculture, water quality monitoring and disaster management. These monitoring tasks are increasingly benefitting from AI techniques, which can automatically extract specific knowledge and insights.
Sentinel-2 captures optical imagery at a resolution of 10 metres per pixel. While this is certainly suitable for several applications, ranging from agriculture (in mainland Europe) to water body monitoring, the images appear very blurry when fine details are needed, such as the exact boundaries of a field, small changes in vegetation or even vehicles or narrow roads.
Super-resolution is the process through which an image can be sharpened and made clearer, such that such fine details are now evident. The goal is to correctly fill in missing details between pixels.
AI techniques are increasingly being applied towards the problem of super-resolution, achieving excellent results in several cases. Nevertheless, it must be said that super-resolution techniques can only be used to improve the image resolution by approximately a factor of four. This means that a Sentinel-2 image could be enhanced from 10m to 2.5m per pixel, enabling smaller features on the ground, such as vehicles, to be distinguishable.
Semablu, a Maltese company, currently delivers satellite imagery with a 2.5m enhanced resolution using its proprietary model, offering a more cost-effective solution than most commercial satellite providers.
Through the PI-EOS (Performance Improvement of Earth Observation monitoring applications using Super-Resolution) project financed by Xjenza Malta, it is successfully demonstrating how super-resolved satellite imagery can be used to improve downstream monitoring tasks which rely on remote-sensing imagery, for applications ranging from change detection to land cover mapping and infrastructure monitoring.
Gianluca Valentino is the principal investigator of the PI-EOS project, which is financed by Xjenza Malta through the Research Excellence Programme. The research team at Semablu also includes Luke Camilleri, Leander Grech, Josef Magri and Glenn Sciortino. For more information about this and other projects, visit www. semablu.com.
A still frame from the National Oceanic and Atmospheric Administration's (NOAA) timelapse of Hurricane Melissa.
* Satellites observe Category 5 Hurricane Melissa, ahead of Jamaica landfall.
After starting monitoring of the new hurricane on October 21, scientists observed as the storm quickly reached the major hurricane class on the Saffir-Simpson Hurricane Wind Scale - in just four days - with sustained winds of 280 kilometres per hour!
For more soundbites, listen to Radio Mocha every Saturday at 7.30pm on Radju Malta and the following Monday at 9pm on Radju Malta 2 https://www.fb.com/ RadioMochaMalta/.
* Satellites don't just take pictures, they take measurements. EO satellites "see" in many parts of the spectrum: visible, infrared, microwave and more. They can measure things we can't see, like plant health (via chlorophyll reflection), sea surface temperature or even soil moisture under vegetation.
* Radar satellites can "see" through clouds and at night. Synthetic Aperture Radar (SAR) satellites (like Sentinel-1 or RADARSAT) emit microwaves that bounce back from Earth's surface - meaning they can observe floods, forest structure or ground deformation day or night, rain or shine.
* EO helps track animals - from space! High-resolution imagery has been used to count elephants, penguins, whales and even identify individual trees in forests - all using AI applied to satellite data.