In order to investigate the possibilities for the generation of digital elevation models with a height uncertainty of 20 mm or less using unmanned aerial vehicles, a survey with a fixed-wing aircraft was carried out over a gravel quarry in Gävle, Sweden. At flying heights of about 80 and 160 m, the resulting imagery had ground sample distances of 24 and 50 mm, respectively. The data was processed in two different software suites – computer-vision-based PhotoScan and photogrammetric RapidStation/RapidTerrain. The results show that PhotoScan was more effective in flat terrain, with height uncertainties of 10 and 17 mm compared with 30 to 40 mm for RapidTerrain, whereas the latter had a clear advantage over undulating terrain.