Optimal Parameters for DEM Extraction from IKONOS data: A case Study of a Coastal Area

Abstract

The gap between aerial and satellite photogrammetry is getting smaller as both systems have benefited from great technology improvements in the last years. However, the major advantage of earth observation satellites continues to be the possibility to acquire data virtually anywhere on the planet without considering border and logistical restrictions and over huge extensions. A pair of stereo images corresponding to a coastal area in the NW of Spain is the basis for this paper, which analyzes certain influential variables in the process of generating a Digital Elevation Model (DEM) from stereoscopic pairs of IKONOS images. The mathematical model, the number of Ground Control Points (GCPs) and their accuracies are analyzed through a sequence of experimental trials. Two main methods are applied for math modeling: a CCRS model (rigorous model) and a Rational Functions model. The influence of the GCPs' positional quality is compared using two different information sources: points measured in 1:5,000 cartography and points surveyed via GPS. The number of GCPs tested varies between 0 and 20. Using different configurations of these variables, 17 models are generated. The best results are achieved with the rigorous model and 16 GCPs measured with GPS with an RMSE of 1.01 m (or an LE95 -Linear Error in Z at 95% confidence level-of 2 m), which is approximately the pixel size of the initial pair. The mathematical model was determined to be the variable with the most influence on accuracy. Moreover, the results suggest that the use of 10-16 GCPs is sufficient. Additional points do not improve the DEM accuracy or may even worsen it in certain cases. Although the study is carried out in a coastal zone, which restricts the GCPs' distribution possibilities, the results are comparable with similar results from inland areas.