Masterarbeiten am Geodätischen Institut

Laufende Masterarbeiten

  • Leveraging uncertainty information to refine semantic segmentation for port and marine structures in the CNN context
    Estimation of the prediction uncertainty of previous trained segmentation networks and introduction of this information to the learning process of a new (modified) model instance to refine damage detection maps.
    Betreuung: Marvin Scherff, Frederic Hake, Hamza Alkhatib
    Bearbeitung: Waqas Munamar
    Jahr: 2024
    Laufzeit: 03/2024 - 09/2024
    © Scherff
  • Strategies for lowering the net land take in Germany and selected countries in the EU
    In Germany, the net land take in 2020 was 54 Hectares per day – mostly used for housing, industry and commerce (excluding extraction areas), public facilities. The German National Sustainability Strategy set the goal to reduce the net land take to 30 Hectares per day in 2020. This political goal was failed by far. The current political goal is to reduce the net land take to below 30 Hectares per day in 2030. Therefore, it is important for local planning authorities (municipalities) to change their strategies in urban development planning (e.g. brownfield development instead of greenfield development). In this thesis, different strategies and approaches for reducing the net land take used for urban development planning should be analysed and compared.
    Betreuung: Jörn Bannert, Winrich Voß
    Bearbeitung: Rahul Chandra
    Jahr: 2024
    Laufzeit: 09/2024 - 03/2025
  • Deep Learning-based Downsampling and Semantic Segmentation of LiDAR Data for Urban Localization of Autonomous Vehicles
    In the context of autonomous driving, localizing a vehicle in an urban environment is challenging due to dense infrastructures that disrupt satellite signals from Global Navigation Satellite Systems like GPS. To overcome this burden, LiDAR sensors can be deployed as they provide precise environmental scans that enhance localization accuracy, but they pose also significant challenges in real-time applications. Therefore, the examined approach in this thesis streamlines two common processes—downsampling and segmenting LiDAR data—into a singular, optimized model using advanced Deep Learning techniques. This innovative framework not only reduces computational load but also enhances the real-time capabilities of vehicle localization systems. The Deep Learning model and its training strategy are steered by a downstream localization filter that evaluates the incorporated processed LiDAR point clouds, ensuring the system's effectiveness in urban navigation.
    Betreuung: Marvin Scherff, Rozhin Moftizadeh, Hamza Alkhatib, Marius Lindauer
    Bearbeitung: Yaxi Wang
    Jahr: 2024
    Laufzeit: 05/2024 – 11/2024
    © GIH
  • Integrated Path Planning in an Extended Kalman Particle Filter Framework for Vehicle Navigation
    Path planning is essential for the efficient and safe navigation of autonomous vehicles in complex environments. This master thesis aims to compare various path planning approaches using a real-world dataset. The study will integrate these algorithms with an advanced particle filter to determine the most effective combination. The ultimate goal is to develop a framework that not only ensures reliable localization but also optimally derives the path for the robot to reach its destination.
    Betreuung: Rozhin Moftizadeh, Hamza Alkhatib
    Bearbeitung: Zao Yin
    Jahr: 2024
    Laufzeit: 01.10.2024 - 01.04.2025
    © GIH
  • Rewetting of raised bogs in Lower Saxony and Baden-Württemberg
    In this thesis, the approach to the rewetting of raised bogs in Lower Saxony and Baden-Württemberg will be examined and the approaches compared.
    Betreuung: Jörn Bannert, Winrich Voß
    Bearbeitung: Amina Alo
    Jahr: 2023
    Laufzeit: 09/2023 - 03/2024

Abgeschlossene Masterarbeiten