Land use change detection using statistical signature matching and rule-based post-processing

This paper presents a method for using multi-temporal TerraSAR-X images together with existing cartographic vector data to identify changes in land use. Typically, this would require existing imagery since direct image comparison is very sensitive for change detection. Here, the goal is to create a system that can reliably identify changes in land use automatically from a small set of multi-temporal radar images, without the necessity of having imagery available from the time of previous interpretation, which in many cases is simply not available. A particular focus is the updating of existing land use maps that were prepared without any use of SAR data.