SPOT 5 HRS geometric performances: Using block adjustment as a key issue to improve quality of DEM generation

摘要

HRS instrument on board of SPOT 5 allows acquisition of stereoscopic pairs in a single pass. Its final purpose is to provide the worldwide database of Digital Terrain Models and Orthoimages, called Reference3D~®, with no use of ground control points and a horizontal location specified as 16 m for 90% of the points. The stake is not only to fulfil the HRS' location specification of 50 m in root mean square but to obtain a location performance as best as possible and to check that the instrument can cope with the realisation of the Reference3D~® database. That's why HRS has been the subject of specific attention even after the end of SPOT 5 in-flight commissioning phase. This comprehensive paper summarizes the steps jointly taken by CNES, IGN, Spot Image and HRS end-users to calibrate, monitor and validate the HRS image accuracy as well as the accuracy of the final DEM products. First, an overview of HRS on SPOT 5 will be given, with a specific attention paid to the instrument geometry and rigorous sensor model available through ancillary data given with the images. Then, methods and means used by the French Space Agency (CNES) and the French Mapping Agency (IGN) to reach the objectives will be described and the final accuracy reached will be given. It concerns: 1. calibration and monitoring of the HRS geometrical parameters, and the several on board improvements brought to the satellite by CNES to improve the location accuracy at the single pair level (i.e. no tie points nor ground control points used); 2. the HRS block adjustment model calibration and the evaluation performed by IGN at the stereo pair level (i.e with tie points used within the pair) and block level (with tie points used between different HRS strips). Finally, the article deals with several comparisons of HRS DEM's against available DEMs (eg. SRTM DTED level 2) and also reviews various evaluations of the HRS and the Reference3D~® products made by independent users against ground truth issued from field measurements or high quality/high scale geodata.