The first surveys have been made with a smart drone with a thermal camera onboard. The drone carried out the mission in autonomous mode (supervised autonomous flight, in accordance with current law dispositions) thanks to onboard computer and installed autopilot, that were equipped with neural networks to master flight and obstacle avoidance.
Mission planning has been composed to guarantee several surveys of the areas of interest at different times, collecting thermal images with a high level of superposition. Those high-quality ortho-mosaics were the first indicators of possible anomalous spill presence in the lake.
After aerial surveys, chemical composition analysis has been conducted. Results showed concentration alterations in some organic compounds such as nitrates.

In the sites where chemical anomalies were detected a further investigation has been assigned to an underwater robot of our own design: Proteo. It could visually identify organic discharge pipes. Unfortunately, as algae and suspended material were high in concentration, obtained data were not significant.
We then moved to satellite image analysis: we normalised them according to inner criteria and elaborated them with Machine Learning algorithms. On this base, we were able to build up a dedicated analysis model, which gave us the ability to precisely spot illegal drain sites. Integrated technology architecture such this one granted H4R the opportunity to win an ESA call with theme “Kick start Artificial Intelligence”.
At the moment the project is being developed.