Data Structuring and Visualization of 3D Urban Data

The major idea of the current research is to build a concept for a 3D GIS appropriate for maintenance and real-time manipulation of municipality data via Internet. A model containing geometric and thematic information about objects, spatial and thematic relationships should be developed. The research focuses the geometric aspect of the problem with emphasis on development of a scheme to exchange data between the database and rendering engines for visualization, navigation and manipulation. The approach for visualization and interaction is based on the utilization of Virtual Reality Modeling Language (VRML) and 3D Formal Data Structure (3D FDS). The VRML is chosen to model the scene for visualization, i.e. to describe geometry, introduce dynamics, illuminate the model, work via Internet, while 3D FDS is considered as basic data structure for data storage and spatial analysis. A variety of visualization queries were performed in order to test suitability of the couple 3D FDS and VRML for interaction with the model via Internet. The results of queries were visualized in HTML and VRML browsers. The experimental work has revealed several shortcomings of 3D FDS with respect to fast composing of the VRML file: 1. The storage of coboundary relationship per face, i.e. left/right body, imposes maintenance of information about “air” body and “underground” body, which speeds down the formation of surfaces and bodies. 2. The explicit storage of boundary relationship per arc, i.e. begin /end node, does not show many advantages but requires extra traversing of one more table for visualization. 3. The implicit description of “holes” complicates the algorithms for coordinates extraction. It can be concluded that all the necessary information for visualization of point, line, surface and simple body objects can be extracted from the 3D FDS. However, organization of data is not efficient to supply data for the followed approach. One direction of the future work is related to extension of the conceptual model toward refinement of the object identification, aggregation of multi-dimensional objects, definition of parameters describing behavior of object, developing a technique supporting LOD visualization. Another aim is refinement of the concept for partitioning of the objects, especially surface objects, in order to avoid rendering pitfalls. The aspects addressed here contribute to clarification of a strategy for real-time manipulation of data.