zurückWo ist die Publikation erschienen?
Master's Thesis (TU Wien)Abstract
Geomorphometry is the science of quantitative analysis of terrain surfaces. By surveying
terrains to quantify their surfaces, it is possible to calculate the geomorphometric
properties, such as heights, curvature, slopes, and distances. These are important for
the analysis of terrains in archaeology, geology, planetary sciences, and others. By using
digital terrain reconstructions, off-site terrain surveying becomes possible. The high
resolution or large scale of terrains are a challenge for real-time rendering at interactive
frame rates for exploration. This requires limiting resolution or loading smaller terrain
parts. The use of terrain streaming allows rendering higher resolution or terrains of
greater extents. As errors remain, it is important to quantify and visualize them.
In this thesis, an out-of-core rendering algorithm for large-scale multi-layered terrain is
presented. The presented streaming algorithm manages to stream scenes with 775 M
triangles and 156 GB on their finest LOD, and a total size of 222 GB, at interactive
frame-rates and on commodity hardware.
Additionally, an improved measurement algorithm for digital terrain surveying of largescale multi-layered terrain is presented in this thesis. The measurement algorithm
using variable-rate subsampling (VRSS) and Shared Edge Detection (SED), is called
VRSS+SED and achieves better results than the fixed-rate subsampling (FRSS) strategy
used in state-of-the-art planetary geology tools such as Planetry Robotics 3D Viewer
(PRo3D). It achieves earlier termination at higher precision for the same number of samples
by intersecting found shared edges with the ray casting plane to analytically calculate
the midpoint between two neighboring primitives. Furthermore, a novel uncertainty
metric called On-Data Ratio (ODR) is presented which allows raising awareness about
the uncertainty in the results of the used state-of-the-art measurements algorithm.
The presented algorithms are evaluated using an implementation in a prototype using
the Unity engine and its Data-Oriented Techstack (DOTS). The algorithms are evaluted
against PRo3D and the results are presented and discussed. The presented implementation
achieves 15x as fast loading times as Pro3D for the 222 GB large scene at a similar
storage size.
Forschungsthemen
Forschungsgruppe