ISBN 978-3-8439-4886-9

978-3-8439-4886-9, Reihe Informatik

Patrick Helber
Machine Learning for Satellite Image Analysis – A Spatial Mapping of Urban Areas

207 Seiten, Dissertation Technische Universität Kaiserslautern (2021), Hardcover, B5

Zusammenfassung / Abstract

In the last decades, the public availability of Earth Observation data grew at an extraordinary rate. Nowadays, satellites, airplanes, and drones monitor our Earth remotely at a unique spatial, spectral, and temporal scale. In particular, satellites play a decisive role in Earth Observation programs. They are aligned to collect data of our entire planet in volumes of terabytes per day and thus record changes on a global scale from space. At the same time, we see remarkable progress in Artificial Intelligence, particularly in Machine Learning, allowing researchers to build models with unprecedented performance. Our society, meanwhile, faces global pressing challenges. An accelerated climate change and an unsustainable development of ever-growing and more densely populated urban areas are two well-known examples. In this context, the interpretation of satellite images enables us to contribute essential information to address respective challenges by tracking changes in Land Use and Land Cover. In this thesis, we study to what extent the advances in Machine Learning models as well as in the spatial and spectral properties of the data taken by novel public satellite constellations can contribute to an enhanced satellite-based mapping. With a focus on urban areas, we perform large-scale studies on a European and a global scale.