Nowadays often a multiplicity of different spatial datasets exists for the same area. In order to use and maintain these datasets effectively they have to be integrated into a multiple representation database (MRDB). In an MRDB, the objects within the different datasets that model the same real-world phenomenon can be linked with each other. When integrating existing datasets these links have to be created by means of automatic matching techniques. In this paper we describe such a matching technique for road data of scales with an order of magnitude difference. The process starts by selecting possible candidates for roads and nodes using a buffer. The candidates are filtered using semantic, geometric and topological information. The remaining node candidates are compared by means of geometric measures, and 1 : 1 links between nodes are created. The node links are converted into road links by a shortest path algorithm. The method has been implemented and successfully tested with data at the scales of 1 : 25.000 and 1 : 200.000.