Science Service System
You are here : Home : Proposals_Summary

Summary of Proposal MTH0382

TitleEnhanced land cover classification using TerraSAR-X data and multiple classifier systems
Investigator Waske, Björn - Freie Universität Berlin, Insitut für Geographische Wissenschaften
Team Member
Dr. Arnason, Kolbeinn - Landmaelingar Islands (National Lad Survey Iceland), N/A
Prof. Hostert, Patrick - Humboldt-Universität zu Berlin, Geomatics Lab, Institute of Geography
Dr. van der Linden, Sebastian - Humboldt-Universität zu Berlin, Geomatics Lab, Insitute of Geography
Prof. Benediktsson, Jón Atli - University of Iceland, Dep. of Electrical and Computer Engineering
Dr. Waske, Björn - University of Iceland, Dep. of Electrical and Computer Engineering
Prof. Chanussot, Jocelyn - Groupe Grenoble INP, Grenoble Institute of Technology
Dr. Fauvel, Mathieu - Groupe Grenoble INP, Grenoble Institute of Technology
SummaryThe primary objective of the study is the application and development of recent classification concepts, which are mainly taken form the field of pattern recognition and machine learning, with a focus on multiple classifier systems and support vector machines. The results in our own studies have underline that these classifier concepts are appropriate for classifying multitemporal and multisensor data sets. The transferability of strategies developed for ERS 2 SAR and ENVISAT ASAR data to TerraSAR-X imagery shall be accomplished within this proposal. A particular focus of the developments is the combination of SAR imagery with multispectral imagery as well as with SAR systems of other band widths (e.g., ALOS PALSAR).The mains research activities can be split in three main sections: (1) Integration of multitemporal information provided by TerraSAR-X, (2) Integration of spatial information provided by TerraSAR-X, (3) Fusion of multisensor data sets. The working package “multitemporal integration” is focusing on the adequate use of multitemporal data, e.g. the determination of robust feature selection and classification strategies to separate cover use classes on a basis of multitemporal data sets. In the module “integration of spatial information” adequate techniques should be developed to derive additional information using spatial high resolution imagery. In this context, multilevel concepts, which are based on the combination of spatial structures at different scales, seems particularly interesting. The final classification strategies should comprise both, the “temporal” and the “spatial” information domain. In the third part of the study, TerraSAR-X data should be combined with additional remote sensing data, as multispectral imagery and data from other SAR sensors. In order to validate the suitability of the developed procedures, test sites within dissimilar environmental settings are considered, including rural and urban areas. Moreover the development of methods for dissimilar regions and applications, will underline the general value of the imagery. The main research will be conducted at the University of Iceland by Björn Waske. Dr. Waske is currently funded by the University of Iceland and an employment until end of 2010 is planned. Additional funds are requested (Icelandic post-doc grant as well as a post-doc grant by the German Research Foundation - DFG).

Back to list of proposals

© DLR 2004-2016