Visual topology in SDI : a data structure for modelling landscape perception

Visual Topology is a used here to describe the spatial relations between objects as they appear in the 2D viewing plane. This thesis sets out the concept, explains why it is needed in Geographic Information Science and suggests how it may be computed through development of prototype software. Section 1 considers the functionality that any Spatial Data Infrastructure would need to encompass in order to support the inclusion of visual analysis into landscape planning and monitoring systems. Section 2 introduces various aspects of visual topology. In particular it sets out how visual intersections of occluding edges may be modelled topologically and formally defines a novel higher level topological structure to the viewing space - the 'Euler Zone' based on the Euler complexity of a graph formed by the occluding horizons in a view. Whether such a graph has meaning to an observer is considered in Section 5, which presents the results of a web ,based forced- choice experiment with significant implications for the role of topology in modelling landscape preference via quantitative metrics derived from 2D maps. Sections 3 and 4 discuss how existing methods for handling perspective models and visualisations need to be improved in order to model visual topology. Section 3 focuses on the limitations of current techniques and design criterion for a new methodology. Section 4 looks at the lessons learnt from developing a prototype implementation (VM-LITE) based on Quad-Edge Delaunay Triangulation, in the VoronoiMagic software package. Some potential applications are highlighted, both within landscape modelling and beyond, before drawing conclusions as to the potential for the concepts and methods respectively. Although important research questions remain, particularly as regards view point dynamics, Visual Topology has the potential to fundamentally change how visual modelling is undertaken in GIS. It allows the analysis of scenes based upon a richer representation of individual experience. It provides the basis for data structures that can support the extraction of generalisable metrics from this rich scene information, taking into account the qualitatively different nature of scene topology as distinct from metrics of shape and colour. In addition new metrics based on attributes only apparent in perspective, such as landform, can be analysed. Finally, it also provides a rationale for reporting units for landscapes with some measure of homogeneity and scale-independence in their scenic properties.EThOS - Electronic Theses Online ServiceGBUnited Kingdo