Matching networks with different levels of detail

This paper deals with the issue of automatically matching networks with different levels of details. We first present why this issue is complex through an analysis of the differences that can be encountered between networks. We also present different criteria, tools and approaches used for network matching. We then propose a matching process, named NetMatcher. This process is a several steps process looking for potential candidates and then analysing them in order to determine the final results. It relies on the comparison of geometrical, attributive, and topological properties of objects. It determines one-to-many links between networks: in particular a node of the less detailed network can be matched to several arcs and nodes forming a complex junction in the most detailed network. An important strength of the process is to self-evaluate its results through the comparison of topological organisation of networks. This paves the way to an interactive editing of the results. The NetMatcher process has been intensively tested on a wide range of actual datasets, thus highlighting its effectiveness as well as its limits.