25.10.2022
For the challenge, machine learning (ML) algorithms were to be developed to determine soil properties from satellite images. This means that parameters important for agriculture, such as potassium content or the pH value of the soil, can be determined directly from hyperspectral images, which saves time-consuming laboratory sampling. The "EagleEyes" team with Frauke Albrecht, Caroline Arnold (DKRZ), Ridvan Salih Kuzu, Kai Konen (DLR) and Roshni Kamath (FZJ) used a so-called random forest algorithm to do this and beat the other 47 teams.
The challenge had been running from 9 February to 31 July 2022 and the winners were announced on 17 October 2022 at the IEEE International Conference on Image Processing (ICIP) in Bordeaux. The teams prize is that their algorithm will be used to analyse image data directly on the Intuition-1 satellite, which is the equivalent of about 50,000 euros. In cooperation with the satellite developers at KP Labs, the ML algorithm will now be made fit for the launch into space planned for 2023.
Further information:
- of the challenge: https://platform.ai4eo.eu/seeing-beyond-the-visible
- of the results by the team EagleEyes: https://github.com/ridvansalihkuzu/hyperview_eagleeyes