DEM generation from satellite data

Digital Elevation Models (DEMs) are required for several tasks like generation of orthoimages, flood planning, erosion control, agriculture, generation of contour lines, visibility check, 3D-views and others. Mainly in developed countries DEMs are available, but not in any case like required. DEMs can be generated by traditional photogrammetry based on aerial photos if they are available and not classified, but also very often more economic by means of space images. Another possibility is the use of airborne laser scanning – this will lead to very detailed and accurate information, but it is expensive. With Interferometric Synthetic Aperture Radar (InSAR) it is also possible to get the height information. The achieved accuracy of DEMs based on space images is mainly depending upon the image resolution, the height-to-base-relation and the image contrast. In addition, systematic image errors of photographic products, but also a limited orientation quality may cause a difference between a relative and an absolute accuracy. Photographic data still do play an important role because of the up to now only a limited number of digital stereo pairs and the very often lower price. Following systems have been analyzed: Metric Camera, Large Format Camera, KFA-1000, MK4, KATE-200, TK-350, CORONA, SPOT, MOMS, IRS-1C/1D, ASTER, IKONOS, QuickBird and SRTM. The manual measurement of DEMs is too time consuming, so most of the data acquisition has to be made by automatic image matching. This includes the disadvantage of a not selected point location. Instead of a DEM, a Digital Surface model (DSM) will be generated, with points located on the visible surface, including vegetation and buildings. The automatic elimination of points not located on the bare ground is possible; corresponding software has been developed and yields to satisfying results. An analysis of DSMs based on InSAR of the SRTM-X-band showed in open areas a relative accuracy of 3m and an absolute accuracy in the range of 6m to 7m. With C-band data the NIMA reached a similar accuracy with the exception of mountainous areas.