Radargrammetric DSM Generation by Semi-Global Matching and Evaluation of Penalty Functions

Radargrammetry is a useful approach to generate Digital Surface Models (DSMs) and an alternative InSAR techniques that are subject temporal or atmospheric decorrelation. Stereo image matching in radargrammetry refers the process of determining homologous points two images. The performance influences final quality DSM used for spatial-temporal analysis landscapes terrain. In SAR matching, local methods commonly but usually produce sparse inaccurate adding ambiguity products; global semi-global seldom applied even though more accurate dense can be yielded. To fill this gap, we propose hierarchical (SGM) pipeline reconstruct DSMs forested mountainous regions using stereo TerraSAR-X addition, three penalty functions were implemented evaluated effectiveness. make accuracy efficiency comparisons between our SGM method method, normalized cross-correlation (NCC) was also same test data. radargrammetric validated against airborne photogrammetric reference compared with NASA’s 30 m SRTM DEM. results show produces height computing exceeds DEM NCC-derived DSMs. function adopting Canny edge detector yields higher vertical precision than other functions. powerful efficient tool high-quality Spaceborne