Student works

The 3DOM unit offers its experience and knowledge to highly motivated visiting students during their BSc, MSc or PhD studies.

Possible topics (contact us - remondino@fbk.eu - for more):

- COLLABORATIVE NAVIGATION: GNSS positioning has been widely used in a large variety of consumer applications, including 3D mapping. The work deals with alternative approaches for positioning and navigation in GNSS-denied situations or environments.  The student should review and investigate other methods, including Signal of Opportunities, Collaborative visual/LiDAR SLAM, UWB, etc.

- VR in MANUFACTURING: The work deals with the use of Virtual Reality solutions - on tablets or smartphones - in manufacturing applications. The candidate should develop a solution able to automatically overlay a digital twin onto its actual/real object and to provide additional information useful to mount, study and understand the object. 

- 3D RECONSTRUCTION AND CLASSIFICATION FROM VERY HIGH-RESOLUTION SATELLITE IMAGERY: The work deals with the use of VHR satellite stereo/triplet images (Pleiades, GeoEye, etc.) for 3D reconstruction purposes and data classification. The candidate is supposed to develop a reliable and powerful methodology to derive accurate and detailed 3D information from VHR images, including semantic 2D/3D data for mapping purposes.

- 3D SOLAR POTENTIAL: The work deals with the estimation of the photovoltaic potential of buildings using 3D point clouds. Common methods are based on raster data while we seek solutions and methods able to handle 3D point clouds in order to consider also facade surfaces. The work includes: State-of-the-art, Development of a method to use building point clouds inn PV potential estimation routines, Development of a tool to visualize the 3D data and results, testing and validation.

- 3D BUILDING MODELING AT REGIONAL SCALE: The work deals with the generation of 3D building models (LOD1/2 resolution) at regional and/or national scale. The procedure to be developed should start from open cartographic data (building footprints) enriched with some attributes. The work includes: State-of-the-art, Development of a method to generate LOD1-2 models from GIS data, Development of a tool to visualize the 3D data on top of DSM/DTM, Testing and validation.

- RE-PHOTOGRAPHY of URBAN SCENARIOS: The work deals with the realization of a (semi- / fully-automated) tool/app for smartphones in order to acquire images in the same location and with the same FOV of an historical image of interest. The work include: Literature review, Development of a smartphone app to read sensors information (GNSS position, accelerometer data, etc.), Develop a methodology to overlap an historical image with the actual view seen by the smartphone camera, Testing and validation.

- 3D SURVEYING OF UNDERGROUND or FORESTRY ENVIRONMENTS WITH SLAM TECHNOLOGY: The topic deals with the use of mono/stereo cameras or LiDAR coupled with SLAM methods to survey underground or forestry scenarios. The candidate should: Review methods, investigate SLAM methods, include SLAM in the 3D reconstruction pipeline, perform tests and analyses in real environments.

- 3D MEASUREMENT, SEGMENTATION AND MODELING OF TREES: The thesis deals with the 3D digitization of trees and their modelling. Starting from 3D point clouds, individual trees should be identified/segmented and then modelled. In particular, the candidate should: Investigate the state-of-the-art of point cloud segmentation, Collect 3D scans of trees, Develop a method to segment individual trees within point clouds, Develop a method to generate 3D models of trees from the segmented point clouds.

- MONOCULAR DEPTH ESTIMATION WITH DEEP LEARNING: The work deals with the generation of depth maps from single images using deep learning methods. CNN or GAN methods should be investigated for single image depth estimation from UAVs/drones or aerial data. The work includes: State-of-the-art, Development / adjustment of neural network for single image depth estimation, Testing and validation using UAVs/drones and aerial datasets.

- AUTOMATIC DIGITALIZATION of (HISTORICAL) MAPS: The thesis wants to investigate how to automatize current manual procedures for the digitalization of urban maps, testing several machine/deep learning algorithms and techniques. In particular, the work should focus on the automatic digitalization of the buildings' footprints in historical maps, as preliminary and necessary data for the 3D/4D city modelling. The candidate should: Explore the state-of-the-art machine/deep learning algorithms, Compare their performances using ground truth data, Develop a digitalization pipeline and test it on several historical maps.

- IMAGE SEMANTIC SEGMENTATION USING DEEP LEARNING ARCHITECTURES: The work deals with the automatic segmentation of images of building facades for the further use of such semantic data inside the 3D reconstruction pipeline. CNN architectures are to be used for this scope in order to extract classes like windows, doors, walls, etc. The work includes: State-of-the-art, Development / adjustment of neural network for image segmentation, Inclusion of such semantic information in the photogrammetric pipeline, Comparative studies and evaluation metrics.

- MEASURING ATTRACTIVENESS OF A LOCATION: The thesis should deal with the collection of touristic (geolocated) images of our region and produce “attractiveness” maps based on density and locations of the shots. In particular, the candidate should: Investigate the state-of-the-art, Collect a large number of images of the most interesting places of the region, Georeference them in case they are not geolocated, Produce attractiveness maps using specific metrics and visualize the results within GIS tools.

- ARCHITECTURES 3D RECONSTRUCTION BY PRIMITIVE FITTING: The thesis work involves the creation of a tool for the generation of semantic 3D models of buildings starting from dense point clouds and architectural primitives. Part of the work could based on the PCL libraries (http://pointclouds.org/) and CAD applications. The candidate should: Create a library with main architectural geometric primitives, Realize a GUI for point cloud loading (ideally web-based), Develop a methodology for best-fitting between primitives and point clouds.

- POSITION AND NAVIGATION: The work tackles the analysis of various positional sensors for geolocalization and moitoring purposes. The working tasks include the analysis of existing solutions and comparison of low-, medium- and high-cost devices/receivers in static and dynamic model. The activities fall within the research project SEC4TD, supported by EIT-RM.

- Clement Fontaine, ENSG Paris, France - Master thesis "Creation d'une Interface pour le Positionnement Automatique des Points d'Appui" (2012)

- Cristina Iacovino, University of Milano,  Italy - Scuola Interateneo di Specializzazione in Beni Culturali - Thesis "L'arco di Augusto a Susa. Studi preliminari per il rilievo e modello 3D" (2012)

- Jose' Miguel Macay Moreira, Gavle University, Sweden - BSc thesis "Using photogrammetric Digital Surface Model in LiDAR software for creating Three Dimensional Buildings" (2012)

- Giulio Stevanato, University of Padua, Italy - Master thesis "Stima del potenziale fotovoltaico su edifici urbani in ambienbte GIS mediante dati LiDAR e fotogrammetrici" (2011)

- Benoit More, ENSG Paris, France - Master thesis "Etude comparative de modelus de creation de MNS sur differents types de terrains" (2011)

- Luigi Barazzetti, Politecnico di Milano, Italy - PhD thesis "Automatic tie point extraction from markerless image block in close-range photogrammetry" (2011)

- Ilaria Trentini, University of Trento, Italy - Master thesis "Approccio multidisciplinare per la documentazione di un sito archeologico: il caso del Riparo Dalmeri" (2011)

- Celine Puig, ENSG Paris, France - BSc thesis "L’utilisation de panoramas pour la modélisation tridimensionnelle" (2010)

- Matteo Rapana', University of Padua, Italy - Master thesis "Modellazione tridimensionale applicata all'analisi archeologica - Il caso studio della chiesa medioevale di S.Martino di Lomaso" (2009)

- Valeria Perlini, CGT Siena, Italy - Master thesis "Fotogrammetria e modellazione 3D applicata all'archeologia - Il caso studio del sito Maya di Copan" (2009)