Autori: S. d'Oleire-Oltmanns, C. Eisank, L. Dragut, L. Schrott, I. Marzolff, T. Blaschke
Editorial: Proceedings of the 4th GEOBIA, May 7-9, 2012 - Rio de Janeiro - Brazil, p.496-500, 2012.
The interest in and need for digital landform mapping is permanently growing, but still lacks fully developed transferable object-based classification approaches. In this study a shared object-based mapping workflow is applied to two different input data sets: aerial photographs and DEMs. We aim to end up with promoting insights leading to a data-independent, so to say transferable, multi-resolution segmentation followed by an object-based classification approach of landforms on different scales. In order to identify a well-suited scale level for data segmentation the Estimation of Scale Parameter (ESP) tool was used. In the next step we developed knowledge-based classification approaches using (border/edge) contrast information for classifying gullies from aerial photographs as well as contextual and terrain layer information as derived from DEMs for mapping drumlins. We found that spectral data (e.g. aerial photographs) as well as terrain data (e.g. DEMs) may be successfully processed. The workflow is furthermore transferable onto reference data sets. Existing differences in contrast/shadowing which occur within aerial photographs from different points in time constitute disadvantages in successful landform mapping which are insignificant when terrain data is used. Knowledge-based identification of landform diagnostic object features improves landform mapping with respect to the input data type and transferability of the classification system. Data integration may enrich the object-based analysis of landforms and will finally advance our understanding of formative processes.
Cuvinte cheie: gully, drumlin, landform mapping, UAV, DEM, multiscale, segmentation, GEOBIA