Abstract
Abstract. Change detection in urban and suburban areas through remote sensing satellite imagery is an important topic. Furthermore, it is of special interest to derive information on the category of detected changes is of special interest. In Boldt et al. (2012), a fullyautomatic change detection method based on a morphological filtered ratio image was presented. This filter step is accomplished by an alternating sequential filter (ASF), which supports the knowledge-driven analysis of the scene. For example, the focus can be set on small-scaled changes caused by vehicles or smaller construction sites. The change detection itself is performed using the automatic threshold method shown in Sahoo et al. (2004) considering the entropies of the fore- and background of the filtered ratio image. In contrast, the presented approach makes use of morphological differential attribute profiles (DAPs) to compare changes detected in high resolution (HR) TerraSAR-X (TSX) amplitude images of Greding (Germany). DAPs are the derivatives of morphological attribute profiles (APs). APs are calculated by applying iteratively attribute openings and/or closings to an input image. Attribute openings (resp. closings) themselves are a combination of connected openings (closings) and trivial openings (closings). DAPs provide the opportunity to derive typical signatures for each pixel of the entire image (Dalla Mura et al., 2010), and, as a consequence, for each detected change segment. Aiming on the categorization of changes, it is shown in this paper that the DAPs represent a promising method for detecting changes with similar semantics automatically.
Highlights
Change detection using multi-temporal high resolution (HR) SAR images is a highly frequented field of research and still a challenging task
The detection itself is done by using the L2-norm (Euclidian distance) as similarity criterion considering the differential attribute profiles (DAPs) signatures calculated for each change segment
Four different attributes are used for the DAP vector calculation: area, std, diag and moi
Summary
Change detection using multi-temporal HR SAR images is a highly frequented field of research and still a challenging task. The high geometrical resolution of, for example, TSX imagery acquired in HR SpotLight mode leads to an incapability of pixel-based approaches, since one pixel represents only a part of an object. This effect is known analyzing optical HR imagery as mentioned in (Weih et al, 2010). In (Boldt et al, 2012), a fully automatic object-based method for change detection in HR SAR image pairs was presented. This method relies on a morphological alternating sequential filtering step that leads to binary change segments after thresholding.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
