Qualification automatique des résultats d'une mise en correspondance de réseaux routiers en vue de la mise à jour cartographique

Herein, we consider the problem of automatically analysing the matching between a road network extracted from a SPOT image and the network given by a cartographic database. This matching relies on two main steps based on a Markovian modeling. The results are then qualified. This provides robust features to improve the registration. Using an iterative scheme, we both improve the registration and the matching between the two networks. The unsupervised results analysis appears to be a precious step towards an automatic map updating.