Verification of digital elevation models from MOMS-2P data

Kurzfassung

The MOMS-2P sensor (Modular Optoelectronic Multispectral Stereo Scanner) is a German stereo camera mounted on the PRIRODA module of the Russian space station MIR. Its main new feature was the along-track three-fold stereo capability. It delivered approx. 65 million km2 of high quality and high resolution (approx. 18m ground resolution) imagery from 1996 to 1999, which is still subject of ongoing image processing to generate digital elevation models and orthoimages throughout the entire world. The three-fold along-track stereo principle and multispectral capabilities of the MOMS-2P sensor were highly advantageous for digital elevation model (DEM) generation compared to other satellite systems in orbit. The majority of all data takes of MOMS-2P were imaged in Mode D, which is a combination of the two inclined stereo and two nadir looking multispectral channels (blue and near-infrared) with a swath width of approx. 100km and a ground resolution of 18m. The data of this mode is well applicable for generating digital elevation models and using the multispectral imagery for creating orthoimages, landuse-classification and further evaluations. This paper describes the latest processing experience in generating high quality digital surface models and orthoimages from MOMS-2P stereo data at DLR. The steps of the processing chain are shortly explained. It is shown that self-calibration via bundle block adjustment is necessary. The quality of the attitude and position data provided by the onboard gyros and the GPS receiver is discussed. It is demonstrated that the highly accurate attitude data is essential for generating precise digital elevation models. The results of an accuracy analysis using digital surface models (DSMs) produced from Mode D image and orientation data from data takes which cover parts of South Germany. The DSMs are checked against a reference DEM with a mesh size of 25m. For a more detailed height accuracy analysis, classification results using the multispectral part of the Mode D imagery allow a discrimination between various landcover types