Latest Tweets




Monitoring of cropland has been critical for several national and international programmes (e.g., Sustainable Development Goals – #2 Zero Hunger, European Common Agriculture Policy). Furthermore, early identification of crops is becoming more stringent in the context of climate change that can influence severely crop yields in some parts of the world. Given the size of te areas to be addressed and the volume of demand, EO based crop monitoirng must increasingly utiliuze AI based approaches. However, cropland classification is a challenging topic because of the constantly changing radiometric signature of crops due to seasons and weather and climatic conditions. This requires the development of a system capable of taking seasonal and weather and climatic variations into account. WIthin the framework of AI based approaches, in order to be economically sustainable, processing costs must also be reasonable.
This project is addressing the entire processing and analysis chain for usiing ML analysis of EO data for crop classificaiton. This includes the identifiaction of which available land cover dataset(s) can provide the best levels of crop information and quality to perform an efficient and conclusive study while meeting specific user needs related to crop monitoring, testing different neural network (NN) configurations, including different input datasets and different approaches to represent data time-series as NN input which are then compared with a baseline classical approach and finally testing different Cloud computing configurations, including the use of the GPU. Beyond the calculation time assessment, this objective will inform on the trade-off between calculation time and platform configuration costs.


Prime contractor