Résumé
The launch of small satellites, CubeSats among others, is skyrocketing easing the development of Earth Observation missions and the implementation of Space applications. At the same time, Artificial Intelligence and deep neural networks algorithms are enjoying impressive success for their results and the diverse applications they enable. The idea of combining nanosatellite standards with Artificial Intelligence capabilities arises thus from a will to push the boundaries of Space systems further. Led by the Grenoble University Space Center (CSUG), the QlevEr Sat mission aims at embedding such an algorithm on board a CubeSat, to process data directly and send only the relevant results to the ground station. From the images taken with the Emerald sensor, forest/non-forest and cloud/non-cloud segmentation maps are inferred with the Teledyne Qormino® QLS1046-Space processor to monitor deforestation. A detailed benchmark of this processor is presented in the paper and confirms its suitability for the mission. The specifically designed pixel-wise classification algorithm leverages the Qormino® processor computing power to produce accurate high resolution binary maps.