Interdisziplinäre Projekte
Wir bieten Interdisziplinäre Projekte (IDP) im Masterstudiengang Informatik an. Wenn Sie sich für Themen im Bereich Fernerkundungsanwendungen interessieren, senden Sie bitte eine individuelle E-Mail-Anfrage, in der Sie Ihr Interesse sowie Ihre relevanten Kenntnisse und Fähigkeiten beschreiben, beispielsweise in den Bereichen Geodatenverarbeitung, Bild- und 3D-Datenverarbeitung, Open-Source-Softwareentwicklung, Research Software Engineering, Programmierkenntnisse insbesondere in Python (Kenntnisse in C++ sind je nach Thema von Vorteil). Die Projekte beinhalten typischerweise (aber nicht ausschließlich) Themen im Zusammenhang mit den Open-Source-Softwareprojekten py4dgeo, CloudCompare und/oder 4D-Visualisierung in unserem Online Potree Viewer. Für formale Details siehe bitte die IDP Webseite / Modul IN2334.
Open Topics in Open-Source Software Development for py4dgeo
We constantly offer open student IDP projects centered around extending, optimizing, and maintaining py4dgeo.
Typical tasks may include:
- Implementing new algorithms or algorithmic extensions.
- Enhancing software architecture, testing frameworks, and continuous integration (CI/CD) pipelines.
- Developing user-friendly tutorials, documentation, or integration tools (e.g., CloudCompare plugins).
Aktuelle Themenangebote für Interdisziplinäre Projekte
Integration of the F2S3 Deep-Learning-Based Deformation Monitoring Method into the Open Python Library py4dgeo
Dense time series of 3D/4D point clouds acquired by modern laser scanning technology are widely used in geosciences, environmental monitoring, infrastructure assessment, and related fields. Reliable spatiotemporal analysis of these point clouds is essential for detecting and quantifying three-dimensional surface changes and deformation. The open-source Python library py4dgeo provides a framework for analysing changes in multi-temporal 3D point clouds, including established methods such as M3C2 and 4D object-based change analysis. Extending the library with additional state-of-the-art methods can provide new possibilities for three-dimensional deformation analysis. F2S3 (Feature to Feature Supervoxel-based Spatial Smoothing) is a deep-learning-based method for estimating 3D displacement vector fields between two point-cloud epochs. It uses learned 3D features to establish point correspondences and a neural network to identify reliable correspondences within spatially coherent supervoxels. Unlike conventional distancebased methods, F2S3 directly provides three-dimensional displacement vectors and can therefore handle displacements that are not aligned with the local surface normal. In this project, the student will study and reproduce the F2S3 method and its existing implementation investigate the architecture of py4dgeo, and integrate F2S3 into the library. The work will include designing the integration, implementing the required interfaces and dependencies, developing tests and examples, and evaluating the resulting implementation on suitable point-cloud datasets.
Reference:
- py4dgeo: Library for change analysis in 4D point clouds.
- Gojcic, Z., Zhou, C., & Wieser, A. (2020). F2S3: Robustified determination of 3D displacement vector fields using deep learning. Journal of Applied Geodesy, 14(2), 177-189.