Quantitative Analysis of
Culling Techniques for Real-time Rendering of Digital Elevation Models
Michael
Hesse
University of Calgary
Department of Computer Science
Calgary, AB, Canada T2N1N4
Marina Gavrilova
University of Calgary
Department of Computer Science
Calgary, AB, Canada T2N1N4
e-mail: marina@cpsc.ucalgary.ca |
http://www.cpsc.ucalgary.ca/~marina |
Keywords
ROAM, culling techniques, error metrics, terrain data, GIS rendering.
Abstract
The paper is concerned with
investigation of effects of culling techniques on quality and smoothness of
terrain data visualization within a 3D interactive environment. We utilize the
Real-Time Optimally Adapting Mesh (ROAM) approach, extended with a number of
efficient techniques such as implicit coordinates method within the
patch array representing ROAM and
the viewpoint dependent triangle rendering method for dynamic level of
detail (LOD) updates. The method, which allows dynamic and interactive first
person view, is combined with the View Frustum culling, the Backface culling and an
innovative Relational
Position Culling technique based on a pre-process point of view approach for realistic terrain
representation and real-time level of detail reduction. Standard error metrics
are used to verify the rendering consistency. The experimentation is conducted
on two data sets representing simple gradual contour changes (Susanville,
California) and complex steep contour changes (Kluane National Park, USA).
Based on the extensive experimentation, the set of heuristics identifying
applicability of each of the culling techniques to a specific terrain data set
is derived.