Microsoft’s SpaceEye Uses AI to “See” Through Cloud Cover

Cloud cover has historically been a persistent problem for Earth observation, given a majority of the planet is covered in fluff. SpaceEye uses artificial intelligence (AI) to produce daily cloud-free optical and multispectral photos of Earth from space. Using the Synthetic Aperture Radar (SAR) instrument from Sentinel-1 (the mission that sent the first phase of the Copernicus Programme satellite constellation to space), SpaceEye creates a base source of cloud-proof radar data. The program then combines the radar data with previous optical imagery to generate an AI-predicted image of what Earth looks like under the pesky clouds.

SpaceEye is intended to benefit the agriculture, land-use monitoring and disaster response verticals, though Microsoft believes its applications could be far-reaching.

Another new Azure Space program, called Project Turing, was announced alongside SpaceEye. Project Turing allows researchers to prepare and enhance geospatial data for better human observation. By combining semantic super-resolution with satellite imagery and improved resolution, the program is able to clarify what would normally be blurry or downright imperceptible snapshots. Project Turing is currently being used to enhance Bing Maps worldwide.

Microsoft originally announced Azure Space last year in an effort to share its Azure suite’s cloud computing capabilities with a broader scientific community. Since then, Azure Space has worked alongside SpaceX Starlink toward bringing high-speed, low-latency satellite broadband for a new Azure Modular Datacenter (MDC), as well as to introduce connectivity to remote areas around the world. According to Microsoft, the MDC has facilitated a unique ability to assist various organizations with the deployment of complete data centers in spaces in which power and building infrastructure are unreliable at best.

Azure Space’s Orbital Emulator has helped satellite developers conduct complex satellite constellation simulations while evaluating and training AI algorithms—all before launching a single satellite. Microsoft states other Azure Space users from the past year have included the U.S Defense and Innovation Unit, Lockheed Martin, and even Land O’ Lakes, which uses the program to determine crop stress, predict yields, and examine the practicality of more sustainable farming methods. (Apparently satellite imagery has more impact on agricultural development than we thought.)