A qualitative model for describing the arrangement of visible cityscape objects from an egocentric viewpoint

摘要

A qualitative model is presented here, which is suitable for describing the relationships between the visible parts of buildings as seen by a street observer. It is intended for use in a Location Based Service (LBS) whereby users access geo-referenced digital datasets on location. Typically such applications filter data according to keywords and two-dimensional spatial reasoning, such as finding all hotels within 500 m. However, a LBS which in addition is able to reason from the user's egocentric viewpoint has the benefit of being able to refer to the arrangement of features in a more natural way, which is particularly useful for dialogue based systems. This research presents a user centred qualitative model which combines and extends previously published projective and topological models. The proposed extensions improve the fidelity of the model by subdividing projective space into finer addressable units, and through their combination the model is able to summarise relationships between complex objects in 3D space, making it suitable for use in queries. The model is demonstrated in a LBS able to establish the visibility of nominated landmarks in a cityscape by using high resolution digital elevation models, which can then support the user who may request information based on the locations of other landmarks (e.g. What's the building on the left of the train station?), or listen to descriptions of landmarks in relation to known features (e.g. the bus stop is in front of the post office). The framework is shown to be able to reason about objects typically in the field of view, and to be suitable for use in spatial queries.