DEM generation based on optical space images

Digital elevation models (DEMs) are a basic component for any GIS, in addition they are required for the generation of the most often generated photogrammetric product, the orthoimages. With the digital elevation models based on the Shuttle Radar Topography Mission (SRTM) in 2000, covering the earth from 56° latitude south up to 60.25° north, height models with a spacing of 3 arcsec (~ 92m) are available free of charge. If such spacing is satisfying the requirements and if the standard deviation of the height in the range of 3m up to 8m can be accepted, no generation of DEMs is justified. Especially in undulated areas the point spacing cannot be accepted, so another source of height information has to be used. Aerial images are not always accessible, but the very high resolution space images can be used. With WorldView-1 a very agile space sensor is available, able to acquire more than one stereo pair over short orbit range. Stereo pairs taken from different orbits with not negligible time difference are causing problems in automatic image matching, but such a problem does not exist with the stereo systems SPOT-5 HRS, Cartosat-1 and ALOS-PRISM. The good results reached with SPOT-5 HRS are under usual conditions only available as height models because the images of SPOT-5 HRS are not distributed. This is different for Cartosat-1 and ALOS-PRISM.
The results achieved with automatic image matching of IKONOS, QuikBird, OrbView-3 as well as with SPOT HRS and Cartosat-1 are shown and compared with the SRTM height models. SPOT HRS showed problems in forest areas; here with Cartosat-1 surprising good matching results have been achieved. An analysis of a SPOT HRS height model, bought as DEM, in the same forest area showed a strange correspondence to the SRTM height model, while the Cartosat-1 DEM was more realistic. The results also have been compared with a DEM from the topographic map 1 : 25000.