Super-Resolution for Land Market

The objective of this activity is to prototype a service offering global Earth coverage of satellite images with approximately 2 m resolution in near-realtime at a reduced cost and shorter revisit intervals with respect to VHR (very high-resolution) commercial optical data. Based on a new scientific approach that enhances ground resolution of satellite images via innovative deep learning methods with GAN (Generative Adversarial Neural Networks), the activity will prototype an automatic solution that will increase Sentinel-2 spatial resolution of several spectral bands (RGB, IR). The method is intended to preserve spectral band properties, allowing the use of any standard EO algorithm such as NDVI, LAI etc. on the resulting super-resolution output images.