Graph-matching model using Gibbsian modeling: application to map/SPOT image road networks for map updating

摘要

Abstract: We consider herein the matching between two graphsrepresenting road networks. This problem is embeddedinto a labeling framework. One graph is taken as areference. A Gibbsian model is proposed to label theother graph. The labels are defined by the noes of thesecond graph. The potentials are defined by the anglebetween the nodes and the length of the associatedfeatures. Therefore, the model is invariant bytranslation and rotation. We apply this model to matcha road network extracted from a SPOT image on the roadnetwork of a cartographic database. This matchingprovides some information for map updating. !14