3 - The Principles of DTM Reconstruction from SAR Images

Abstract

A more or less refined knowledge of Earth’s varied terrain is necessary for many thematic applications of satellite imagery. Since the advent of the SPOT family of optical satellites (in 1986), many studies have been conducted to achieve digital terrain models (DTMs) that have been used in a number of topographical and hydrographical applications such as erosion studies, geographic information systems (GIS), etc. To establish the DTM, specific techniques for acquisition of optical images have been developed by changing the direction of optical image acquisition through agility sensors: left/right for SPOT 1–4 and front/rear for SPOT 5 and ASTER. Alongside these achievements with optical data, approaches based on synthetic aperture radar (SAR) image data quickly showed the potential of this type of imaging. Indeed, varied terrain and SAR imaging are inseparable because as a result of their fundamental principle of side-looking, radar images are sensitive to the terrain. If these effects can hamper the interpretation of an image in the eyes of a photointerpreter, it follows that they are very useful for assessing terrain. With the emergence of SAR sensors (SIR-A and B, ERS), various techniques for terrain reconstitution (radargrammetry, radarclinometry, interferometry) have been implemented and have shown that SAR could be the ideal tool for reconstructing the Earth’s landscape. This is also a result of the work that the Jet Propulsion Laboratory (JPL) developed in 2000 for the Shuttle Radar Topography Mission (SRTM), which established an initial global DTM of nearly 80% of the land emerged from the DTED-1 level (mesh of about 90m × 90m). Then in 2010, the DLR (German Aerospace Center) sent the Tandem-X satellite (TerraSAR-X add-on for Digital Elevation Measurement) into orbit and together with the TerraSAR-X Mission, it established the DTM, named WorldDEM, which has been available since 2014 on a mesh of 12m × 12m, with an absolute accuracy of 10m and a relative accuracy of 2m (marketed by Infoterra GmbH) of 80% of the land emerged from the DTED-1 level (mesh of about 90m × 90m).