A hierarchical graph matching method to assess accuracy of network extraction from DTM

More and more elevation data and methods are available to automatically map hydrographic or thalweg networks. However, there are few methods to assess the network quality. The most used method to compare an extracted network to a reference network gives global quality information on only geographic criterion. The method proposed in this paper allows a network assessment compared to a reference network whose results can be interpreted more easily and more related to networks morphologies. This method is based on a hierarchical node matching within a graph. Nodes are classified by hierarchical level according to their importance in the tree-structured network. Then, a matching process seeks for nodes pairs between the two networks based on the geographic distance. The hierarchy introduces a priority order in the matching. The relative location of nodes pairs is checked in order to ensure a topological consistency. Finally, similarity statistics based on nodes matching counts are computed. While the usual method only takes into account a geographic criterion, the presented method integrates geographic, geometric and topologic criteria. It is an interactive and object-byobject matching. Moreover, the hierarchical approach helps comparing networks represented at different scales. It provides global statistics but also step-by-step maps that helps characterizing the spatial distribution of network delineation errors.