Generation of 3D building models from archived city area maps by fusing information from multiple sources

• Main Authors: Dong, Chaoqun, Martel, Roman, Chen, Kan, Johan, Henry, Erdt, Marius
• Other Authors: -
• Format: Working Paper
• Language: English
• Published: Nanyang Technological University 2021
• Subjects: Engineering::Computer science and engineering::Computing methodologies::Image processing and computer vision; Engineering::Computer science and engineering::Computing methodologies::Simulation and modeling
• Online Access: https://hdl.handle.net/10356/145920
• Institution: Nanyang Technological University

This paper proposes a pipeline for the automatic generation of 3D building models for city areas based on information from archived city area maps. These maps are typically created and archived by city authorities for several decades. As such, they exhibit several challenging properties like a mixture of handwritings and typewriter font styles, varying layouts and notation standards, low contrast and physical damages. To tackle these challenges, we propose to extract and fuse information from multiple sources. In the proposed pipeline, we firstly locate and extract text content within the city area maps to obtain the essential information to identify described buildings. Secondly, based on this information, we retrieve the building height information and addresses from a public housing database. Then, we extract the building shape and size information on the basis of the obtained addresses through an online map API. Lastly, utilizing all the acquired building information, we generate 3D models of the buildings and their neighborhoods in the CityGML LOD1 format. The whole pipeline and its individual components are tested on a dataset of city area maps provided by city authorities.