Master Thesis »Obtaining altitude road profiles by using super-resolution imaging«

In Fraunhofer ITWM, in the division Mathematics for Vehicle Engineering, we work on methods for modelling and analysis of usage variability and reliability in vehicle engineering. A special focus lies on the statistical description of vehicle usage models based on geo-referenced data, planning and evaluating measurement campaigns, as well as the extrapolation of data to design loads. We are developing and implementing new modelling and simulation methods and we are applying these methods in projects with leading industry partners.

High-quality elevation profiles of roads are important for simulating vehicle loads and energy consumption. They are difficult to derive from freely available data sources such as Shuttle Radar Topography Mission (SRTM) digital elevation models (DEMs). This is due to the relatively low resolution of 1 arcsecond by 1 arcsecond, which corresponds to about 30 m by 30 m, since the typical road width is much smaller. In addition, roads are man-made constructions that can mark discontinuities in the environment, such as when roads run on bridges.

Currently, we solve these problems using complex interpolation methods and manual processing of special situations. The idea in this project is to develop automatic algorithms that increase the resolution of freely available DEMs by using super-resolution imaging. To this end, we aim to combine high-resolution land use information from freely available road network data, such as the Open Streetmap data, and merge this information with the information from the DEMs. Promising methods here are machine learning techniques, such as deep neural networks, since we can evaluate a large pool of high-resolution (1 m x 1 m) DEMs for specific regions in Germany.