Photogrammetric Processing of a TerraSAR-X Image: Evaluation of Products

摘要

TerraSAR-X was the first from a series of very high resolution SAR satellites, which have significantly enlarged the range of cartographic applications of SAR data. For such applications the use of photogrammetric processing, either independently or combined with optical images became meaningful, In this paper the procedures that may be applied on a single SAR image are developed and the quality and accuracy of their products are investigated regarding the impact of using various models or changing the parameters of external data (GCPs, DTM). A Spotlight mode TerraSAR-X image is used for a case study. The area is a mountainous terrain, of 5×10km size, located in the north-eastern region of Athens, Greece. Three alternative registration/georeferencing methods are investigated: the Physical Sensor Model (Range-Doppler method), the DLT and the 2nd order RPF. Orthoimage production using four different DTMs of various quality and density are investigated. Geometric accuracy and quality are investigated using Independent Check Points and through an analysis of the slopes and aspects of the terrain in combination with the geometry of the sensor. Automatic techniques for the extraction of the road network are used; however the results are not satisfactory. In addition, a manual extraction of the roads using visual interpretation is made and the results are evaluated in terms of completeness, correctness and quality. According to the results derived from the above procedures, the use of high resolution TerraSAR-X images is very promising for a variety of applications for which medium scale geometric accuracies are required.