Historical 3D city models have been increasingly used for the preservation and communication of the cultural heritage to a wider and more diversified public. In the recent years, they have also been of a growing interest in other domains such as in urbanism or in economy. However, their potential for supporting new use cases has been restricted by the difficulty to generate these models. Historical 3D city models can only be reconstructed from historical sources, such as historical maps, and this means dealing with all sorts of constraints and inaccuracies. As a result, reconstructing historical 3D city models is a challenging process that is to date still essentially manual and time-consuming. This thesis investigates to what extent the reconstruction of historical 3D city models can be automated. Several existing methods for extracting building footprints from historical maps have been tested and compared so as to identify the pros, cons and use cases of each method and all the challenges of working with historical maps. Based on these experiments a fully automated methodology was developed. It relies on three main stages: (1) the processing of the historical maps to extract the building plots, (2) the subdivision of these building plots into individual building footprints and (3) the reconstruction of a LoD2 historical 3D city model using 3D procedural modelling. This methodology was implemented with historical maps from two different study areas, Delft and Brussels, and for different epochs in order to reconstruct a dynamic historical 3D city model for these cities. The results show that the methodology workflow developed in this thesis allows to reconstruct automatically historical 3D city models for different historical maps collections and for different study areas. The main differences between the two case studies, Delft and Brussels, regard the implementation details (i.e. data availability, running time and user-defined parameters) but similar results are obtained, which show the suitability of the methodology to be applied for other study areas. Two elements are identified as main factors influencing the quality of the results obtained: the quality of the scanning process and the symbology of the historical maps. For historical maps that were properly scanned, with sufficient spatial resolution and strict symbology rules, the methodology provides accurate results by identifying more than 84% of the building plots in the ground truth and classifying properly more than 89% of the building plots. In addition, all historical 3D city models reconstructed have their geometries valid at more than 99%. Overall, this thesis provides a methodology for reconstructing automatically historical 3D city models from historical maps along with guidance and hints about this process and about a series of other methods, so that any user can find the most suitable method for their needs. All source codes and data of this thesis are available at https://github.com/camilleMorlighem/histo3d.

The Deployment of Indoor Point Clouds for Firefighting Strategy project was realised as a Synthesis Project of the Geomatics Master Programme of the Built Environment Faculty at the Technical University of Delft. This project was executed by a team of five Master students in collaboration with the Dutch response team collective Veiligheidsregio Rotterdam-Rijnmond. The objective of this project is to develop an information system that makes use of indoor data to support tactical decision-making during fire emergency responses. The main challenge that response teams are facing when they develop deployment plans is the lack of appropriate information about indoor spaces. As a result, response teams may end up relying on inaccurate assumptions which can lead to dangerous situations. New technologies such as SLAM devices and augmented reality displays, combined with processing techniques, can be used to supply them with the information needed to make the right choices. The result of this project is a prototypical information system containing an interactive, 3D environment that can receive updates, merge data from different data sources, and accommodate mixed reality information sharing in real-time.