Building and roof type classification on 3D city models

Sitong Li, Chengzhi Rao, Chi Zhang, Wei Wei

3DBAG is a 3D building modelling reconstruction tool that aims to keep the 3D building models current with the latest building stock, roof shapes, and elevation information available. The complete potential of 3D building models can be harnessed and used in numerous practical scenarios by making use of current 3D models and different geo-information datasets. 

In collaboration with Spotr, a high-tech company specializing in digital building inspection using geospatial data, our primary objective is to propose a novel method for classifying roof and building archetypes. By utilizing the 3D models from 3DBAG, satellite images, and other related geospatial data, this method aims to integrate machine learning and topology-based approaches and explore the 2D and 3D features of various buildings seamlessly.

Real-world buildings can be highly intricate and challenging to classify, the ideal vision of this method is to extend its scope beyond residential buildings to encompass public and commercial structures, and effectively identify archetypes within complex roof combinations and arrangements. The results derived from this research hold substantial importance in various domains, including insurance estimation, energy consumption management, solar panel installation, disaster evaluation and post-processing applications. This project is poised to unlock the full potential of 3D building models, offering a wide array of practical real-world applications.    

More information