Gradient-Based Surface Reconstruction and the Application to Wind Waves

New gradient-based surface reconstruction techniques are presented: regularized least absolute deviations based methods using common discrete differential operators, and spline based methods. All new methods are formulated in the same mathematical framework as convex optimization problems and can handle non-rectangular domains. For the spline based methods, either common P-splines or P1-splines can be used. Extensive reconstruction error analysis shows that the new P1-spline based method is superior to conventional methods in the case of gradient fields corrupted with outliers. In the analysis, both spline based methods provide the lowest reconstruction errors for reconstructions from incomplete gradient fields. Furthermore, the pre-processing of gradient fields is investigated. Median filter pre-processing offers a computationally efficient method that is robust to outliers. After the reconstruction error analysis, selected reconstruction methods are applied to imaging slope gauge data measured in the wind-wave facility Aeolotron in Heidelberg. Using newly developed segmentation methods, it is possible to detect different coordinate system orientations of gradient field data and reconstruction algorithms. In addition, the use of a zero slope correction for reconstructions from the provided imaging slope gauge data is justified. The impact of light refracting bubbles on reconstructions from this data is part of this thesis as well. Finally, some water surface reconstructions for measurement conditions with different fetch lengths at the same wind speed in the Aeolotron are shown.

Level thesis:degree_level

master

Grantor dc:publisher

Universität Heidelberg

Year

2017

Author dc:creator

• Scherr, Tim

Contributors dc:contributor

• Jähne, Bernd

Repository record source_url

http://www.ub.uni-heidelberg.de/archiv/23653

OAI identifier oai:identifier

oai:archiv.ub.uni-heidelberg.de:23653