Doctoral Defense in Computational Engineering - Dirk Norbert Baker

Doctoral candidate:
Dirk Norbert Baker

Title of thesis:
Design and evaluation of data analysis and augmentation approaches for functional-structural plant models

Opponents:
Dr. Michael Pound, Professor of Computer Science at the University of Nottingham, UK
Dr. Thomas Odaker, Researcher at the Leibniz Supercomputing Centre, Germany

Advisor:
Dr. Morris Riedel, Professor of Computer Science at the University of Iceland

Other members of the doctoral committee:
Dr. Ebba Þóra Hvannberg, Professor of Computer Science at the University of Iceland
Dr. Hanno Scharr, Researcher at the Forschungszentrum Jülich, Germany
Jens Henrik Göbbert, Researcher at the Forschungszentrum Jülich, Germany

Chair of Ceremony:
Dr. Rúnar Unnþórsson, Professor and Head of the Faculty of Industrial Engineering, Mechanical Engineering and Computer Science, University of Iceland

Abstract:
Analyzing plant data for precision and robust agriculture is a key component of adapting to climate change. Plant data is multi-modal, experiments are directly restricted by plant growth, and the data analysis is often supported by image capturing. Particularly in the case of functional-structural plant models, the understanding of experimental data, the primary indicators for plant development and health, is crucial to calibrating robust and representative models of the real world. However, experimental data is also limited in its scope, which makes it more crucial that the data analysis pipeline is not only robust against noise but also yields as much value to researchers as possible. The focus of this thesis is the embedding of plant models into a virtual environment in such a way that it is informed by experimental data and compatible with both a more precise extraction pipeline and a more scalable data generation pipeline. Synthetic data generation is a key aspect of this thesis, as it is one of the promising ways to combat data scarcity for biological data analysis. This thesis contributes to the field by establishing a data generation framework that is compatible with modern high-performance computing systems and is based on a realtime communication standard. The virtual world embedding of the plant models also yields the possibility of measuring in-place, allowing the coupling of plant models to visualization to exhibit true digital twin behavior. Within the virtual embedding, data extraction is also more precise, allowing users to directly interact with challenging data sets to increase the precision and robustness compared to traditional methods. The output of this PhD thesis is a distributed synthetic data-based training framework called Synavis, full realistic virtual scenes that contain functional information and are computed scalably, and a data analysis pipeline called VRoot that directly assists researchers in their data analysis whenever automated approaches need human intervention.

About the doctoral candidate:
Dirk Norbert Baker is a PhD student in Computational Engineering at the University of Iceland. He completed a Bachelor's degree in Scientific Programming and an apprenticeship in Software Engineering in a dual study program. After his studies, he worked as a software engineer in the Virtual Reality group at RWTH Aachen University while completing a Master's degree in Applied Mathematics at the University of Applied Sciences Aachen, Germany. He started his PhD at the University of Iceland in 2021. During his PhD, he has conducted research in the fields of plant sciences, computer vision, and high-performance computing at the Jülich Supercomputing Centre in Germany. Dirk is currently working as a Research Software Engineer at the Helmholtz Centre for Environmental Research (UFZ).