A STRUCTURE-ORIENTED MATCHING APPROACH FOR THE INTEGRATION OF DIFFERENT ROAD NETWORKS
M. Zhang, L. Meng, H. Qian
Department of Cartography, Technical University Munich, Germany
The widespread GIS applications require an intensified study of geo-information from different sources covering the same geographic space. Data matching is one of the fundamental techniques that help make different datasets interoperable. In the past decades a lot of algorithms focusing on how this matching can be done have been developed for various purposes, such as (1) increasing the applicability of the existing data by transferring attributes or object classes from one data set to another, (2) evaluating and improving the data quality by comparison of different datasets, (3) maintaining or updating datasets in MRDB (Multiple Representation Database), and (4) providing navigation solutions for LBS (Location Based Service). A majority of these matching algorithms have revealed satisfied matching performance on certain data types in selected test areas. However, the problem of uncertain matching remains either in areas where the context conditions are too complicated (like around highways) or when one of the datasets contains little or no meaningful semantic information at all. In some sense a street network can be regarded as one unit constituted by various road structures, such as single carriageways (centre line of road), dual-carriageways (parallel lines), roundabouts, narrow passages, slip roads around cloverleaf junctions etc. Since the various road structures take on quite different geometrical or topological characteristics from each other, it is hardly possible to efficiently match all of them using the same criteria or methods. Keeping these unfavourable conditions in mind, the authors present a structure-oriented approach for the matching between different street networks. The approach is characterized by three consecutive steps: structure recognition, process modeling, and process execution. Structure recognition aims at identifying and describing typical object clusters, i.e. road structures based on their spatial and/or semantic characteristics. During the process modeling, different structure categories resulted from the recognition trigger different matching algorithms as well as the necessary criteria. Process execution as the final step takes place to operate the matching approach. With regard to the general problem of searching scope and computing intensity, the authors put forward a reasonable execution sequence for various structure categories. The paper will particularly demonstrate the matching strategies for complicated road structures.
The proposed matching approach has been successfully implemented for the integration of the road layers from Basis DLM (Basic Digital Landscape Model of German mapping agencies) and Tele Atlas. Due to its generic nature in handling incomplete geometric and semantic information that is available in the datasets to be matched, the approach reveals a high flexibility and allows an extension to include further structure categories.