Treffer: Digital Urban Twins for heavy rain events - An open source QGIS plugin for machine learning classification of residential buildings using CityGML with additional datasets.

Extreme weather events such as heavy rains are an increasing challenge. The potential impact of flooding on residential buildings can be simulated using digital twins. However, when using geometric-semantic information from diverse data resources, such as 3D city models, zoning or cadastre, the data must be carefully selected and programmatically prepared for the simulation. In this study, we present how a use-case driven classification was generated for the residential buildings in the city of Dresden, which is used to estimate the damage potential. The research focuses on both the supervised building classification with a neural network and the open source software framework. Data management is done with the 3DCityDB in PostgreSQL. QGIS is used for visualisation and user interaction. The Python-plugin automatically classifies more than 70,000 residential buildings based on 37 residential building classes. The hierarchical classification is challenging due to the ground truth sample size of about 21,000 and the heterogeneous distribution of the samples. The core of the method is the training and validation utilising random forest as machine learning method. With the developed toolset, classification results can be visually checked in a subsequent step using QGIS. Additionally, the classification, might be corrected manually for individual buildings using mobile mapping data, if necessary. Eventually, the assigned classes are fed back into the official CityGML city model as a new attribute, enabling a realistic damage potential analysis, in a free and publicly available 3D-WebGIS platform. The project is funded under the Smart Cities pilot programme of Germany. [ABSTRACT FROM AUTHOR]

Copyright of ISPRS Annals of Photogrammetry, Remote Sensing & Spatial Information Sciences is the property of Copernicus Gesellschaft mbH and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)