Knowledge-based quality assessment and improvement of digital elevation models from interferometric SAR data

Abstract

We present a knowledge-based approach to assess the quality of DEMs. Inputs are the DEM derived from single-pass interferometry in X-band, the coherency of the interferometric pair and L-band data for land-cover classification with respect to basic height classes. Our approach uses techniques from approximate reasoning to deal with vague height information and ambiguous class assignments. The results are twofold. 1) We provide a confidence map, giving for each pixel a confidence value for the DEM. 2) Regions with low reliability (smooth surfaces, shadows) are replaced by an appropriate interpolation.