Automated image matching between geocoded Landsat-TM scenes and MOMS-2P stereo imager for DEM and orthoimage production

Abstract

The German along-track stereo scanner MOMS-2P delivered 3-line stereo imagery of many parts in the world from 1996 till 1999. Its satellite platform was the Russian space station Mir. The relative accuracy of the exterior orientation of the MOMS camera based upon GPS and gyro measurements (1 m and 1-2 arcseconds, respectively) was sufficient with respect to MOMS ground pixel size of 17 m. In order to get the absolute orientation, bias and drift parameters and some values of the interior orientation of the camera have to be estimated via bundle adjustment. For this task, ground control points (GCP) have to be provided. For large areas in Afghanistan, Jordan, and Libya, geocoded thematic mapper imagery was available. Automated image matching is used to extract thousands of GCP even though the time gap between Landsat and MOMS imaging ranges from 8 to 11 years. The geoid height of the GCP is taken from available digital elevation models (DEM) of lower accuracy. The GCP enter a bundle adjustment which results in improved values of exterior and interior orientation. These are then used in DEM derivation and orthoimage production for the three MOMS viewing directions. DLR's MOMS stereo workstation software is used for all these tasks. The fit between TM and MOMS orthoimages and between the orthoimages of the off-nadir looking MOMS channels is checked via automated image matching. Mean and standard deviations of the shifts are found to be in sub-pixel range.