Opportunities and challenges of on-board AI-based image recognition for small satellite Earth observation missions

Abstract

The satellite industry is rapidly growing. There has been a significant increase in the number of new small satellites that are launched, which is complemented by the rapid pace of the development of image recognition algorithms. Convolutional neural networks (CNNs) in particular, have achieved state-of-the-art performance in computer vision related applications. Combining both and running an AI algorithm on-board the satellite to observe and recognize any natural disaster directly from the orbit is an important opportunity. This paper presents notable challenges that are generally involved in an Earth Observation small satellite mission and further challenges that are posed by combining it with AI-based image recognition on-board the satellite. This study discusses an approach that is feasible mainly for a fleet of small satellites.