Available Master Theses
Master’s thesis topics offered by Design++ Faculty Network. Contact us to propose a new one.
Master thesis topics:
- Download Synthetic form-based colouring of point cloud models using AI and LiDAR intensity and classification fields (PDF, 100 KB), Supervisor: Philipp Urech
- Acceptance study of glacier cover, Supervisors: Philipp Urech, Adrienne Grêt-Regamey, Daniel Farinotti
- Download Co-Creation Tool for Potential Areas for Wind Energy Exploitation (PDF, 133 KB), Supervisor: Ulrike Wissen
- Download Validity of Virtual Reality Landscapes: Eye Tracking Analysis (PDF, 210 KB), Supervisors: Ulrike Wissen, Laura Schalbetter
- Download Communicating Urban Streetscape Qualities: Audiovisual Simulation with Unity and Sumo (PDF, 213 KB), Supervisor: Ulrike Wissen
- Download Audiovisual Simulation of Streetscapes (PDF, 172 KB), Supervisor: Ulrike Wissen, Somea Desarzens
- Download Form-based computation of mean temperature change with LiDAR and AI (PDF, 256 KB), Supervisor: Philipp Urech
- Download Synthetic point cloud generation of Swiss landscape typologies on topography using AI and LiDAR open data from Swisstopo (PDF, 146 KB), Supervisor: Philipp Urech
Synthetic form-based colouring of point cloud models using AI and LiDAR intensity and classification fields.
Laser-scanning technology, also LiDAR, creates precise geometric documentation of the landscape. The LiDAR sensor can detect the intensity returned by the laser, but it cannot capture the colour spectrum. For this reason, a coloured point cloud model needs to be coloured by projecting coloured images possibly in the same direction of the sensor (i.e., vertically for airborne data or spherically for data recorded with a terrestrial laser scanner).
The image projection causes major issues resulting in images difficult to understand or visually misleading. These problems may derive from insufficient calibration (due to movement or geolocation mismatch) and multi-layering (the foreground image is projected on a wrong background object; e.g., canopy is projected on the ground).
The task of this master thesis is to develop a systematic approach using deep learning to color lidar datasets according to formal features. The nation-wide dataset by Swisstopo provides a variety of landscape forms where specific colorings could ease the comprehension. The goal is not to create photorealistic 3D models, but rather find a representation style that conveys clarity and information.
More information
Planning of Landscape and Urban Systems PLUS
Dr. Philipp Urech
Synthetic point cloud generation of Swiss landscape typologies on topography using AI and LiDAR open data from Swisstopo
Landscape patterns are the result of cultural and natural processes that create sets of vegetation and settlement patterns that are related to the underlaying topographic shape. These patterns are documented in 3D by LiDAR data, which represents the geometric structure of landscape typologies in terms of shape, grain, density, alignment, height, vertical structure and so on. Above-ground landscape patterns have a deep connection with the topography on which they are standing. By harnessing generative capabilities of machine learning, it would be possible to produce quickly and accurately fictional scenarios according to a given topography.
The task of this master thesis is to invent a workflow capable to generate fictional landscape patterns on a provided topography. The generation can be keyword-based and can use semantics from other databases (such as othophotos) to create a semantic library for landscape patterns represented as LiDAR point cloud models. The aim is to create a generator of synthetic point cloud models on a digital terrain model (DTM). This research will open the doors to new questions on landscape morphology and its typological blueprint.
Planning of Landscape and Urban Systems PLUS
Dr. Philipp Urech
Form-based computation of mean temperature change with LiDAR and AI
The physical form of landscape features such as groves, settlements and road infrastructure has a direct impact on the average temperature of their surroundings, either increasing or decreasing it. Precise morphological information on terrain and land cover has enabled a paradigm shift in various environmental disciplines. Understanding the physical form of the environment could lead to a better understanding of ecological, water and climatic variables. This understanding can strengthen tasks in landscape design and planning to merge functional needs, aesthetic and cultural values, and mitigate risks such as flooding, urban heat, noise and air pollution. The development of simple tools for planners and designers will support operations aimed at shaping the territory with all its idiosyncratic qualities. Thanks to precise computer algorithms, the geometric documentation produced by the surveys could provide valuable clues as to how the man-made landscape functions.
The task of this master thesis is to correlate form, type, materiality of landscape features with temperature maps using Artificial Intelligence. The form, type and materiality will be drawn from open LiDAR data provided by Swisstopo that carries coordinates (XYZ at a planar resolution of about 20 points/m2), basic classifications (ground, vegetation, buildings, water), and intensity values that can help distinguish materiality. The temperature maps shall be taken from high resolution climatic analysis available as open data such as the bioclimatic map of Geneva (2020). The aim is to train a classifier using AI, so that it can map and visualize the change in mean temperature solely from LiDAR data.
Planning of Landscape and Urban Systems PLUS
Dr. Philipp Urech