Abstract: Automatically deriving multiple representations of a road network at different levels of detail is desirable for various geospatial applications. This paper focuses on deriving simplified representation of complex road junctions from its detailed representation. It is based on the observation that a road junction is a designed functional structure that consists of functional elements. Each type of element usually has a shape pattern, while the composition of elements has a structural pattern. A road junction can therefore be represented by means of structural description and recognized by means of structural pattern recognition. The structural patterns of road junctions are represented as directed attributed relational graph (DARG) in this study. The collection of common road junction patterns constitutes a set of graph templates to be matched to. In order to simplify road junction representation, a road network is first converted to an attributed graph. Then, junction patterns are searched in the resulting attributed relational graph of road network. That is a process of subgraph matching. Ullman’s algorithm for subgraph matching is adopted in this study. Once a junction is recognized, it can be simplified according some predefined method. Experiments have been carried out to evaluate the proposed technique. It is shown that the proposed technique is quite effective in describing and recognizing road junctions.