Automatic Quality Assessment of Gis Road Data Using Aerial Imagery - a Comparison between Bayesian and Evidential Reasoning

This paper describes the framework for automatic quality assessment of existing geo-spatial data. The necessary reference information is derived from up-to-date digital aerial images via automatic object extraction. The focus is on roads, as these are amongst the most frequently changing objects in the landscape. In contrast to existing approaches for quality control of road data, a common and consistent modeling and processing of the road data to be assessed and the road objects extracted from the images is carried out. A geometric-topologic relationship model for the roads and their surroundings is set up. The surrounding context objects (for example rows of trees, or rows of buildings) support the quality assessment of road vector data as they may explain gaps in the extracted road network. Algorithms are defined for the evaluation of existing relations between extracted objects and the database road objects and thus quality measures are yielded. Mostly, more than one extracted object gives evidence regarding one database object. Therefore, the gained quality measures have to be combined in order to reach an overall quality value for the respective object. In the present work two approaches are used for this reasoning and are compared: a probabilistic one and an approach based on the Dempster-Shafer-Theory of Evidence. Results carried out on real and simulated data show that the overall approach is both reliable and efficient. Both models for the reasoning have major differences, however, differences between the results from both approaches only show up in some cases.