GEOSCAN Search Results: Fastlink

GEOSCAN Menu


TitleEvaluation of Radargrammetric DEM from RADARSAT Images in High Relief Areas
DownloadDownloads (Preprint)
 
LicencePlease note the adoption of the Open Government Licence - Canada supersedes any previous licences.
AuthorToutin, Th
SourceIEEE Transactions on Geoscience and Remote Sensing (Institute of Electrical and Electronics Engineers) vol. 38, issue 2, 2000 p. 782-789, https://doi.org/10.1109/36.842007
Year2000
Alt SeriesEarth Sciences Sector, Contribution Series 20042820
PublisherInstitute of Electrical and Electronics Engineers (IEEE)
Documentserial
Lang.English
Mediapaper; on-line; digital
File formatpdf
Subjectsdigital terrain modelling; remote sensing; incident angles; RADARSAT; SAR imagery; Ground Control Points (GCP); radargrammetry; stereo images; radiometric differences; stereoscopy vs monoscopy
AbstractThe paper addresses digital elevation mode (DEM) generation with RADARSAT synthetic aperture radar (SAR) fine mode images (F2-F4 and F1-F5) over high relief areas. Shallow look angles and small intersection angles were chosen to reduce layover and radiometric differences in the stereo-images, and thus increase the performances of the image matching. The radargrammetric DEMs are thus evaluated as a function of processing methods (stereo viewing and editing) and different geometric parameters (intersection angle and derivative topographic surfaces). The first results show that the ground control points (GCPs) have to be acquired with stereoscopic viewing instead of monoscopic viewing to avoid more than 30% error increase in the DEM. An error of one pixel for the image matching has been estimated. The DEMs have also to be edited with semi-automatic and stereo-viewing to increase their coherence and consistency, mainly at mountain ridges and lakes. The results also show that a 50% decrease in the intersection angle (8° versus 4°) resulted only in 15% decrease in accuracy (24 m versus 28 m), in contradiction with the theory. Finally, the DEM accuracy is almost-linearly correlated with the terrain slopes but not with the terrain aspect (except better results for the foreslopes).
GEOSCAN ID219622

 
Date modified: