Diploma and Master Theses (authored and supervised):
"A Composable and Reusable Photogrammetric Reconstruction Library";
Supervisor: W. Purgathofer, S. Maierhofer;
final examination: 2018-03-22.
hotogrammetry is the act of recovering information from photographs. Thanks to digital photography and cheap consumer cameras, the recent years have seen a rising interest in photogrammetry as a more flexible alternative to traditional surveying tech-nologies, for example laser scanners. Specifically, the flexibility of photo cameras makes photogrammetry suitable for outdoor scenes unfavourable for heavy equipment, such as construction sites or drone flights. However, commercial photogrammetry software often makes experimentation with novel techniques diÿcult, since the subject is mathematically complex and code is often diÿcult to modify or closed source.
In this diploma thesis, we present a composable, reusable photogrammetry library which aims to facilitate confident experimentation in a scientific context. We derive important components needed for photogrammetry, including the representation of data (Feature Extraction and Feature Matching), obtaining three dimensional scene structure from two dimensional images (Pose Recovery) and computing a globally consistent model from arbitrarily many images (Structure From Motion and Bundle Adjustment). Adhering to Functional Programming paradigms, we compose these basic components to form more complex modules, and show how to build an extensible and flexible library API on multiple levels of abstraction. Accompanying code listings and examples are included, showcasing example reconstructions with our library.
Electronic version of the publication:
Created from the Publication Database of the Vienna University of Technology.