The aim of this project is to establish an automatic pixel classification service from Sentinel 2 products at global scale. This classification is based on bottom of atmosphere radiometric information and its goal is to help Sentinel 2 users to create downstream services. Up to now, ESA provides Sentinel 2 level 1 products (top of atmosphere reflectance) without any interpretation. Thus, each company which wants to develop a specific service has to construct a full radiometric process line in order to generate specific informations from its own algorithms. This hinders the use of EO data from company without radiometric scientific expertise and more broadly without EO expertise.
This project will customize a dedicated capabiltuy to classify individual pixels using an existing ML methology. The output of this project is a web service for Sentinel 2 users which provides, on demand and in close real time, an interpretation map of any area on the planet. This interpretation will be based on the ground information only, and will help them to target their desired use-case. It will eliminate clouds, and ignore atmospheric effects.