Research on high resolution synthetic aperture radar image urban scene classification based on local semantic representation

This paper proposes a method to classify urban scenes using high resolution synthetic aperture radar image based on local semantic representation. In this method, urban scenes are represented by a middle-level feature, called local semantic category histogram, which is used to overcome the semantic gap between low-level features and high-level user semantics. By a simple maximum likelihood classifier, we discriminate four different urban scene categories: the first class of residential areas, the second class of residential areas, the third class of residential areas, and the nonresidential areas. The method is evaluated on a real high resolution TerraSAR-X image. The experimental result has shown that the local semantic category histogram can well represent urban scene.