A semantic zoom approach in GIS to rule-based landscape visualization | |
| Author | Chamnan Kumsap |
| Call Number | AIT Diss. no. RS-05-2 |
| Subject(s) | Geographic information systems Landscape Visualization |
| Note | A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Technical Science |
| Publisher | Asian Institute of Technology |
| Abstract | High-realism landscape visualization of an almost photographic or ultra-realistic nature can be achieved by using techniques such as texture mapping or planting growth models of vegetation upon a terrain surface. In so doing, composing three-dimensional scenes for landscape visualization is a very time-consuming and resource-greedy process and becomes increasingly uncontrollable as a landscape to model consists of a huge number of vegetations and architectural structures. At close-range landscape visualization realistic 3D models of objects at high level of detail are needed to enhance scene realism. Lots of 3D object models lead to a clumsy hardware performance during scene rendering. In fact, objects further away are subject to less attention and poorer visibility. They should be rendered with low level of detail to save render time. That calls for a technically and practically sound method of scene simplification in which an appropriate level of detail is allocated to 3D objects for rendering 3D scenes within a shorter time. Apart from the scene-composing difficulties, the information visualized from the scenes is subjective and abstracted to a large extent unless there is a tool to help describe the landscape under visualization. The author has established rules in GIS for allocating a proper level of detail to 3D object models before integration to the terrain for scene rendering. The rules that drew on geo-spatial and non-spatial relations of GIS database namely Visual attention, Vertical projection, Horizontal projection, Visibility decay, and Selective rule were proposed. Visual attention rule stemmed from the density and domination of the objects that revealed the objects' significance upon the terrain. The projected image size i.e., height and width, of projected 3D models on the scene was employed to originate geometric Vertical and Horizontal projection rules. The visual interaction between 3D models and an observation point was mathematically modeled to an exponential decay function and the rule was named Visibility decay. It described the way that vegetation visibility decayed as the viewing distance increased. However, each rule was established to address the significance of the models under a particular relation of the models in the landscape. Selective rule was designed to generalize the performance of the rules with the purpose of enhancing scene's visual realism and reducing render time. Eventually, a rule-based scene was composed and simplified by Selective rule before rendering. The study shows the visual and statistical comparison of the rule-based and highest resolution scenes. A significant decrease was observed in either the number of computed polygons or rendering time of the rule-based scene. The rules performed their level-of-detail allocation task very adequately and Selective rule was able to select the proper results that met the requirement of scene simplification. Thus, the rules can be applied to generate any forested landscapes according to their advantages found in this research. In addition, semantic zoom tools were designed and implemented in GIS for rule-based landscape visualization. The concept and implementation of the semantic zoom technique was assessed by questionnaire. The respondents evaluated the semantic zoom tools in terms of their ease of use and additional knowledge of the landscape. The result turned out that most respondents favored the tools because the knowledge of the landscape was additionally provided. The findings inferred that the semantic zoom tool was another novel approach that not only provided landscape viewers a better understanding but also brought about a good visual perception of the landscape. The result can be applied to the design of built environments. |
| Year | 2005 |
| Type | Dissertation |
| School | School of Advanced Technologies (SAT) |
| Department | Department of Information and Communications Technologies (DICT) |
| Academic Program/FoS | Remote Sensing (RS) |
| Chairperson(s) | Borne, Frederic; |
| Examination Committee(s) | Honda, Kiyoshi;Hoang Le Tien;Barczi, Jean Francois;Mackaness, William A.; |
| Scholarship Donor(s) | Royal Thai Air Force; |
| Degree | Thesis (Ph.D.) - Asian Institute of Technology, 2005 |