OVERWATCH is developing a new set of machine-learning algorithms to map floods, fires, and damaged areas. The algorithms will be designed to use both Copernicus Sentinel as well as drone data. Our team diversity and expertise will complement the model development and training.
In particular, for Sentinel-2, super-resolution techniques will be trained on the 10m (visible and NIR bands) and the 20m (red edge, SWIR bands) to enhance the spatial resolution of the source data up to 4 times (5m) to understand the effectiveness of such an approach for producing more accurate maps. The trained algorithms will be evaluated and validated in real cases, namely floods and fires that will occur within the project duration, while their operational capabilities will be assessed with drones.
Pre-event image: SPOT6 © Airbus DS (2022), (acquired on 01/03/2022 10:40 UTC), GSD 1.5 m, approx. 0% cloud coverage in AoI, 16.2° off-nadir angle), provided under COPERNICUS by the European Union and ESA, all rights reserved.
Post-event image: SPOT6 © Airbus DS (2022), (acquired on 15/09/2022 10:15 UTC, GSD 1.5 m, approx. 0% cloud coverage in AoI, 34.5° off-nadir angle), provided under COPERNICUS by the European Union and ESA, all rights reserved.
Reference: Copernicus Emergency Management Service.
Project funded in the Horizon Europe Programme (CALL:HORIZON-EUSPA-2021-SPACE) under Grant Agreement 101082320
