[Back]


Talks and Poster Presentations (with Proceedings-Entry):

M. Attwenger, P. Dorninger, F. Scholten, G. Neukum et al.:
"Fusion of HRSC and MOLA data for high quality Mars DTM computation";
Poster: 1st Mars Express Science Conference, Noordwijk, The Netherlands; 2005-02-21 - 2005-02-25; in: "Abstract Book", (2005), 132.



English abstract:
Digital terrain models (DTMs) derived from High Resolution Stereo Camera (HRSC) images by area-based matching exhibit some deficiencies. The method's dependency on albedo features of the surface results in noisy or even pointless regions. Furthermore, point matching methods are not able to directly reconstruct linear structural features such as sharp terrain edges or crater rims. Therefore, the resulting surface models may show unnaturally rough areas caused by measurement noise, may have more or less large gaps where image features are not detectable, and terrain discontinuities appear smoothed.

Our approach overcomes these problems and improves the quality and subsequently the interpretability of the final DTMs. The main ideas are aimed at eliminating the influence of the measurement noise by using point classification, bridging pointless areas with already available topographic information (e.g. MOLA, Mars Orbiter Laser Altimeter), and improving the modeling of linear structures with the help of an adequate structure line detection algorithm.

The classification method is applied iteratively starting from an initial surface model. First, a MOLA DTM was used as initial value. But, it turned out that a coarse DTM derived from the original HRSC point cloud can be used as well. Therefore, regions with sufficient HRSC point distribution are derived from these points only, whereas pointless areas are bridged by MOLA data.

The method has been tested in different areas. The mean measurement noise can be reduced by a factor of two from about 43 m to 24 m. It could be shown that features not discernable in the original point cloud stand out clearly in the DTM derived from the classified points. The presented approach is expected to provide the basis for improving the quality of subsequent products, such as orthoimages, maps, and results of specific investigations.

Created from the Publication Database of the Vienna University of Technology.