Proceedings Article | 23 October 2014
KEYWORDS: Synthetic aperture radar, Image segmentation, Geographic information systems, Amplifiers, Amplitude modulation, Curium, Feature extraction, Sensors, Microwave radiation, Satellites
Change detection based on remote sensing imagery is a topic highly on demand with various fields of application. Probably, disaster management is the best known, where it is crucial to get fast and reliable results to enable a suitable supply of the affected region. Another important issue, for example in city or land-use planning, is the regular monitoring of specific regions of interest. For both scenarios, it would be significant to have information about the type or category of the detected changes. Since High-Resolution (HR) Synthetic Aperture Radar (SAR) is in opposite to optical sensors an active technique, it is well-capable for all change detection topics where a regular monitoring is intended. SAR sensors illuminate the investigated scene by their own microwave radiation and most applied microwave wavelengths make SAR nearly independent from atmospheric effects like dust, fog, and clouds. Moreover, the time of day makes no difference using SAR sensors. Acquired in HR SpotLight mode 300 (HS300) by the German satellite TerraSAR-X (TSX), images have a resolution of better than one meter, which allows to separate small objects placed close together. In this paper, a concept of change analysis focusing on small-sized areas is presented. Those change areas can be caused by man-made objects (e.g. vehicles, small construction sites) or natural events like phenologically based changes of the vegetation. Since the presented change analysis concept deals with the analysis of time series imagery, other seasonal also man-made caused changes (e.g. agriculture) can be detected. Furthermore, the concept comprises the categorization of the detected changes, which separates it from many of the existing change detection approaches. It includes five central components given by the change detection itself, the pre-categorization of change pixels, the feature extraction for change blobs, the analysis of their spatial context, and the final decision making forming a categorization statement. In all steps, Object-Based Image Analysis (OBIA) methods are utilized. As test area, the airport of Stuttgart (GER) and its surroundings containing heterogeneous change categories is considered. At current state, one time series consisting of 11 HS300 amplitude images acquired in ascending (ASC) orbit direction is available. For the evaluation of results, several reference data are useable comprising optical satellite, terrestrial information and GIS vector data.
