Science Service System

Summary of Proposal LAN1639

TitleLand Cover and Change Detection Mapping Using TerraSAR-X
Investigator Balzter, Heiko - University of Leicester, Geography
Team Member
Dr Tansey, Kevin - University of Leicester, Department of Geography
Mr Spies, Bernard - University of Leicester, Department of Geography
Mr Van Beijma, Sybrand - University of Leicester, Department of Geography
Mr Kose, Mustafa - University of Leicester, Department of Geography
SummaryTerraSAR-X has proved to be a significant source in the provision of high resolution land cover mapping, particularly in cloudy areas of the world. However, little research has looked at the application of the data over African biomes, particularly with regards to seasonal variability which is a feature of many African regions. Rain clouds often obscure test sites during the rainy season leaving much information to be inferred from coarse scale optical imagery. TerraSAR-X could add significant information during these times, a feature of particular interest to the humanitarian aid community, as well as for REDD+ applications and land cover mapping in general. Furthermore, the transferability and robustness of SAR classification algorithms when applied to different test sites and biomes has not been considered in detail. Therefore, this work will firstly consider the use of multi-polarised TerraSAR-X imagery and multi-frequency SAR for land cover mapping over different biomes in Africa (Semi-arid and Forested). Secondly The TerraSAR-X archive provides a useful source of information about coastal habitats. Even though initial studies have been carried out successfully to identify different coastal vegetation habitats with SAR, no focus has been put on the temporal variations and vegetation structure dynamics within these habitats. Thirdly it is proposed to use the TerraSAR-X archive to research and identify buildings in urban areas and to detect environmental change with multi-polarimetric and multi-frequency SAR. Monitoring and identification of different land cover types such as bare soil, forested areas, coastal zones and urban areas are essential in a changing world and can assist in situations where various land use conflicts arise. All the selected sites overlap with archive data from ENVISAT ASAR (dual polarised) and ALOS PALSAR (dual or quad polarised). The intention is therefore to also request the C-band and L-band data from ESA via a Category-1 Proposal over the same test sites, to develop a multi-frequency classification algorithm for different seasons of semi-arid and forested regions of Africa, temporal dynamics of coastal zones, and multi-frequency land cover classification surrounding urban areas. Analysis will be performed by processing and calibrating the SAR imagery using Envi, Gamma and NEST software, to derive the sigma-nought images, inter-channel decomposition and coherence images. After this SAR signatures will be extracted for various land cover types (urban / settlements, water, bare, forest, dense vegetation, sparse vegetation, coastal vegetation) for each of the study sites. Texture measures will be calculated. These signatures will be used to develop a transferable classification algorithm over forested and semi-arid regions of Africa. Classification algorithms that will be tested are Random Forests, Support Vector Machines and object-based image analysis. Other classification algorithms will also be considered as the research progresses. Peer-reviewed journal article publications, conference papers and at the end of the project, this will be a major contribution towards the investigators’ PhD theses.

Back to list of proposals

© DLR 2004-2016