Aiteanu, Fabian: Extracting and Exploiting Structure in Point Clouds for Editing and Interpolation. - Bonn, 2022. - Dissertation, Rheinische Friedrich-Wilhelms-Universität Bonn.
Online-Ausgabe in bonndoc: https://nbn-resolving.org/urn:nbn:de:hbz:5-68949
@phdthesis{handle:20.500.11811/10540,
urn: https://nbn-resolving.org/urn:nbn:de:hbz:5-68949,
author = {{Fabian Aiteanu}},
title = {Extracting and Exploiting Structure in Point Clouds for Editing and Interpolation},
school = {Rheinische Friedrich-Wilhelms-Universität Bonn},
year = 2022,
month = dec,

note = {Modern movies and computer games employ rich virtual environments to offer the consumer an immersive experience. Digital models are often easier and more cost-efficient to create than acquiring real-world film footage of exotic locations, items, creatures or plants. With advances in range scanning technology, acquisition times for point clouds have been reduced to the order of seconds. Point clouds offer a very detailed representation of 3D data, but at the same time pose considerable challenges due to the amount and nature of data.
In this thesis, we present our recent approaches to working with point clouds as a representation of 3D data and corresponding editing operations. First, we demonstrate how to apply non-linear editing operations on a point cloud based 3D model, allowing to bend, strech, rotate or scale parts of it in real-time, while ensuring smooth deformations and intuitive user interactions. A key ingredient is the use of quaternions to describe and compute point neighborhood relationships in a pose-invariant manner. Second, we introduce a technique to compute the skeleton of a biological tree, given real-world laser scan data. Such data may contain densely sampled regions, such as the tree trunk, while branches in the tree crown may be sampled sparsely and suffer from occlusions and noise. Our approach can determine the positions of branches and bifurcation points and also compute branch radii.
Finally, we combine the previously described methods to create a framework for generating tree models by interpolating and extrapolating from given tree samples. The input tree point clouds are skeletonized and then transformed into a quaternion-based representation in a shape space. Different points in that shape space represent different trees, which are similar to the input samples, but still distinct and unique, so that they can be employed to populate diversified virtual environments.},

url = {https://hdl.handle.net/20.500.11811/10540}
}

Die folgenden Nutzungsbestimmungen sind mit dieser Ressource verbunden:

InCopyright