Nowadays cartographic objects of different resolutions are hold in different coexisting databases. This implies an extensive amount of work for updating an object in all resolutions. One way to reduce this is to build a multi-resolution database which holds and links different representations of the real-world objects and allows to automatically pass updates to all linked representations (requiring algorithms for propagation of updates). In order to ensure the autonomy of applications on the local databases, we propose the architecture of a federated database for such a multi-resolution database. A main requirement for setting up a multi-resolution database is to "identify" the different representations of real-world objects. Therefore we propose a multistage procedure as follows: - Semantic classification: Identify the sets of objects to compare with one another - Compute geometrically possible matchings within these classes - Postprocessing: Automatically select correct matchings from the set of possible matchings by applying prepared rules - Manual select correct matchings for the remaining possible matchings, which were not automatically detected as correct or incorrect We present a framework for the needed semantic classification, a concept for rule-based selection as well as an algorithm to compute possible matchings. We have enhanced the formerly known buffer growing algorithm for computation of possible matchings and implemented it in PL/SQL for use in spatial databases based on Oracle9i (with the spatial data cartridge). The enclosing object matching process is supported by a graphical user interface utilizing stored database procedures for the mentioned steps and rules. © 2004 International Society for Photogrammetry and Remote Sensing. All rights reserved.