An attempt to derive a DEM and ortho-photo map from hyperspectral data in the high resolution mode (Casi/SWIR)

摘要

Future Remote Sensing data will include hyperspectral data more and more. This study frames in the preliminary studies to investigate the possibilities of these types of data. In this framework experimental flights were organised by VITO (Flemish Institute for Technological Research) and OSTC (Federal Office for Scientific, Technical and Cultural Affairs) on 13th of September 2002 with a Dornier 228 aircraft carrying the CASI/SWIR sensor, over different test sites in Belgium. The flights were done by NERC and VITO and framed in the STEREO-program of OSTC. This paper deals with the "Ghent test site" over a peri-urban area, and focuses on the 3D-geometry of these data. These hyperspectral data are characterised by relief-displacement and differences in relief. These displacements occur in a direction away from the centre of the image. For remotely sensed images this is the direction away from the flight-line. In the monoscopic analysis of such images, these displacements will disturb the interpretation, the classification and the mapping of the image information. These distortions can be compensated for with a DEM. Very often this DEM has to come from an external data source such as laser scanning or aerial photography. However, when hyperspectral data is taken stereo-scopically, it is possible to derive a DEM from these stereoscopic hyperspectral images, and so avoiding the cost to purchase an external DEM. To obtain stereoscopic imagery, two images were taken with an overlap of 80% in the so called "metrical mode" i.e. with a high spatial resolution (0.5m) and a lower spectral resolution. A third image was taken in "spectral mode", for interpretation issues. A DEM and orthophoto were generated from the hyperspectral data by means of digital photogrammetry. A number of Ground Control Points were collected for absolute orientation, by means of DGPS. The geometric precision of the DEM and orthophoto derived from these data are evaluated using DEM and orthophoto derived from aerial pictures on different scales.