|M.Sc Student||Eliana Bshouty|
|Subject||Updating OpenStreetMap using LoD1 Building Data Extracted|
from Contributed Photographs
|Department||Department of Civil and Environmental Engineering||Supervisor||Dr. Dalyot Sagi|
|Full Thesis text|
3D city models are valuable and useful information for a wide range of applications, including facility management, emergency response and rescue operations, architecture, city planning - to name a few. Due to the large impact of 3D city models on the one side, and the high cost of generating these digital infrastructures on the other, automatic and fast procedures for the generation of different grained 3D city models are required. Perhaps the most important features in 3D city models are building features, since they serve as a major geospatial element in most environmental applications. These are mainly produced by mapping agencies and companies that use high-end technologies, such as laser scanning and photogrammetry.
This research investigates the use of user-contributed data for this task, exploring automatic computational ways and methods for reliable construction of 3D building models. To this end, analytical algorithms were developed and analyzed that use data from OpenStreetMap mapping infrastructure, containing building footprints, together with on-ground building photographs. This included algorithms related to numerical optimization, image processing, data mining and 3D reconstruction. This research developed two approaches for the task of generating accurate and detailed LoD1 building models, which can also serve as a consistent solution for updating missing height information in geographic databases, such as OpenStreetMap.