Improving Quality of Digital Elevation Models Derived from Satellite Stereo Images Using Geospatial Techniques

Abstract

AbstractDeriving Digital Elevation Models (DEMs) from satellite stereo images are the standard in geospatial studies world over, as it is pragmatic, practically feasible, economically viable and technically sound. Specific satellite missions were put and stereo data has been acquired in these missions. Techniques have been developed globally by many institutes for the utilization of this data for image matching and DEM generation. During the last couple of decades, automatic DEM generation has been taken up at global and regional levels, SRTM global DEM, ASTER GDEM and CartoDEM being some examples. The quality of a DEM measures how accurate the elevation is at each pixel (absolute accuracy) and how accurately the morphology is presented (relative accuracy). However, the DEMs produced automatically from satellite stereo images suffer from the problems of sinks/spikes, non-uniformity of elevation over water bodies, ill-conditioned along streams and rivers. The quality of DEM could be improved either at the process level or at the product level. There, will not be any manual intervention in the automatic DEM production chain and hence has to be dealt with at the product level. The quality of such DEMs can be improved interactively by applying custom tools/methods/algorithms. Multi-temporal satellite stereo data comes in handy to improve absolute accuracy and produce DEMs at super resolutions (which may not be feasible with single-date data). This article discusses methods of spike/sinks correction, hydrological conditioning, combining multi-temporal DEMs into a single DEM, and also DEM production at super resolutions.KeywordsDigital elevation modelStereo imageSRTMASTER GDEMCartoDEM