The document presents a novel method for automatically detecting and reconstructing building radar footprints from single very high resolution synthetic aperture radar (SAR) images. Unlike most existing methods, the proposed approach can be applied to single images. It extracts low-level features from images and uses a production system to organize them into more structured primitives. It then introduces the concept of semantic meaning to represent the probability that a primitive belongs to a certain scattering class, like double bounce or roof. This semantic meaning helps compensate for the lack of detectable features in single images by allowing selection of the most reliable primitives and footprint hypotheses. The method demonstrates high overall detection rates on a test TerraSAR-X image, reconstructing flat- and g