Railway parameter extraction with high-precision UAV-photogrammetry: a feasibility study
Abstract. This paper presents a feasibility study on the extraction of railway parameters from high-precision UAV photogrammetry. A two-stage methodology is adopted: first, point-cloud accuracy and rail-surface coverage are optimized through controlled laboratory and field acquisitions; second, the resulting point clouds are evaluated against the local precision requirements for railway parameter extraction. Laboratory tests show that the SfM workflow is intrinsically capable of sub-millimeter agreement under controlled conditions. In the field, standard nadir and higher-altitude flights proved insufficient to reconstruct the narrow rail-side geometry required for accurate gauge estimation. The best results were obtained by combining a 15 m nadir flight with additional side-looking images at 2.5 m and 5 m, yielding local fitted-surface RMS values down to 0.25 mm and about 0.37 mm on the rail running surface. Although the best lightweight UAV configuration still showed a 2.2 mm gauge discrepancy relative to TLS, the laboratory validation and the strong performance of optimized acquisitions indicate that tolerance-compliant railway parameter extraction may be achievable with higher-end UAV platforms and improved rail-side visibility. Overall, the results confirm the strong potential of UAV photogrammetry for near-industrial railway documentation.