| The amount of available data in a Geographical Information System (GIS) is growing at a very high rate. Many of the data sets are at different scales, requiring cross-referencing and integration. Real-time data processing and display capabilities of such systems are also at an all-time high, making generalization a feasible and necessary concept for interactive and automated procedures. Digital Terrain Models (DTM) are now a significant component of GIS, thus requiring generalization. An extensive literature review of DTMs—from basic definitions and conceptual frameworks to available algorithms for generalization and different applications—shows that there is no complete solution at the moment due to the lack of analytical theory underlying the generalization procedures.; This research explores the possibilities of combining theory with practice, that is combining the generalization of the underlying structure (SLM) and other terrain structures with the heuristic methods used to generate visually pleasing views using selective filtering. An algorithm based on the Constrained Delaunay Triangulation has been developed and thoroughly tested, using visual analysis and statistical measures. The results proved that the unification of model generalization and statistical generalization is both feasible and desirable. Several methods for comparing the outcome of generalization procedures have been developed, allowing an objective insight into the underlying structure of the terrain and theoretical aspects that lead to a better understanding to automated generalization procedures. |
