A Hybrid Method for Deriving Dtms from Urban Dems

This study focuses on the automatic extraction of DTMs from urban DEMs produced by image correlation. Many methods have been proposed in the literature yet processing correlation DEMs in any kind of areas remains a challenge. This paper presents a hybrid approach that combines complementary aspects of both TIN-based and segmentation-based techniques. Unlike previous work involving two complementary modules, the two techniques closely interact during the process. The DEM is first segmented and classified into ground and aboveground regions using contextual information. A DTM is then derived from the ground regions using a TIN-based technique. The classification and the DTM estimation are finally iteratively performed until stability. The hybrid approach for DTM extraction has been tested over several representative data sets and compared with the TIN-based and regionbased approaches applied independently. It clearly shows that coupling complementary approaches improves the quality of the resulting DTM, as well in dense urban areas as in rural or hilly areas.