ThetaSpace develops and maintains the CloudlessEO software, an artificial intelligence program based on deep neural networks that fuses Synthetic Aperture Radar (SAR - Sentinel-1) satellite data and cloud-corrupted Optical satellite imagery (Sentinel-2) to remove clouds and their shadows and fill those gaps with geometrical and spectral information as if there were no clouds!
How it works
CloudlessEO accurately detects and masks clouds on every Sentinel-2 imagery, then exploits the temporally closest sentinel-1 in two processes:
CloudlessEO uses multiple inputs to accurately generate cloud-free optical images that match the ground truth, such as geolocation, prior image acquisition, etc.
Thetaspace is currently developing a beta version of CloudlessEO that aims to provide on demand cloud-free, high resolution optical satellite imagery for the RGBNIR (Red, Green, Blue and Near Infrared) spectral bands with a global coverage of Earth's land surface with a revisit time of up to 6 days.
Although SAR, thanks to its longer wavelength, can penetrate through clouds and can deliver data day or night, it has many limitations that prevent it from being an alternative to optical imagery. SAR is a complex data that is much more difficult to interpret than an optical image. SAR requires tedious preprocessing , which requires the use of special software and its knowledge. Unlike SAR, optical imagery is user friendly and can be immediately interpreted by virtually anyone.
We believe that our solution CloudlessEO, efficiently overcome the cloud obstruction problem in optical imagery and: