RESULTS OF EXPERIMENTS ON AUTOMATED MATCHING OF NETWORKS AT DIFFERENT SCALES

Many geographical databases of the same area are produced and maintained. In order to remove inconsistencies between those databases, and in order to facilitate the updating process, a close integration is required. This paper reports the results of experiments of data matching between the networks of two IGN databases at different scales (road, electric, hydrographical, railway and hiking routes networks). We illustrate the main results of these experiments through four different aspects. The first aspect is the identification of data that can not be matched because they only appear in one database. Some of these differences between the contents of the databases are clearly explained by the specifications: they reflect the difference between points of views. Some other differences are just discovered in the data: they reflect the different sources used to build databases and inconsistencies due to errors. The second aspect is the analysis of differences and inconsistencies between databases when corresponding objects are identified. In particular, we identified differences between attribute values, geometric descriptions, but also topological relationships between objects. A third studied aspect is the degree of automation of the matching process. We managed to automatically match from 90% to 100% of objects in networks, depending of the complexity of the networks. Finally, the fourth considered aspect concern the interactive checking of results. We identified ergonomic difficulty to visualize data and results of matching. We thus propose some solutions to overcome these difficulties.