| Topographic eminences are defined as upwardly rising, convex shaped topographic landforms that are noticeably distinct in their immediate surroundings. As opposed to everyday objects, the properties of a topographic eminence are dependent not only on how it is conceptualized, but is also intrinsically related to its spatial extent and its relative location in the landscape. In this thesis, a system for automated detection, delineation and characterization of topographic eminences based on an analysis of digital elevation models is proposed. Research has shown that conceptualization of eminences (and other landforms) is linked to the cultural and linguistic backgrounds of people. However, the perception of stimuli from our physical environment is not subject to cultural or linguistic bias. Hence, perceptually salient morphological and spatial properties of the natural landscape can form the basis for generically applicable detection and delineation of topographic eminences. Six principles of cognitive eminence modeling are introduced to develop the philosophical foundation of this research regarding eminence delineation and characterization. The first step in delineating eminences is to automatically detect their presence within digital elevation models. This is achieved by the use of quantitative geomorphometric parameters (e.g., elevation, slope and curvature) and qualitative geomorphometric features (e.g., peaks, passes, pits, ridgelines, and valley lines). The process of eminence delineation follows that of eminence detection. It is posited that eminences may be perceived either as monolithic terrain objects, or as composites of morphological parts (e.g., top, bottom, slope). Individual eminences may also simultaneously be conceived as comprising larger, higher order eminence complexes (e.g., mountain ranges). Multiple algorithms are presented for the delineation of simple and complex eminences, and the morphological parts of eminences. The proposed eminence detection and delineation methods are amenable to intuitive parameterization such that they can easily capture the multitude of eminence conceptualizations that people develop due to differences in terrain type and cultural and linguistic backgrounds. Eminence delineation is an important step in object based modeling of the natural landscape. However, mere 'geocoding' of eminences is not sufficient for modeling how people intuitively perceive and reason about eminences. Therefore, a comprehensive eminence parameterization system for characterizing the perceptual properties of eminences is also proposed in this thesis. Over 40 parameters are suggested for measuring the commonly perceived properties of eminences: size, shape, topology, proximity, and visibility. The proposed parameters describe different aspects of naive eminence perception. Quantitative analysis of eminence parameters using cluster analysis, confirms that not only can eminences be parameterized as individual terrain objects, but that eminence (dis)similarities can be exploited to develop intuitive eminence classification systems. Eminence parameters are also shown to be essential for exploring the relationships between extracted eminences and natural language terms (e.g., hill, mount, mountain, peak) used commonly to refer to different types of eminences. The results from this research confirm that object based modeling of the landscape is not only useful for terrain information system design, but is also essential for understanding how people commonly conceptualize their observations of and interactions with the natural landscape. |
