Abstract Buildings represent fundamental elements of urban systems, and the management and updating of their representations in spatial information systems are therefore essential. Despite long-term research efforts, the automatic detection and extraction of building footprints from available data sources remain challenging. Difficulties arise from factors such as vegetation cover, close proximity of buildings in dense urban areas, varying levels of detail, inconsistent building definitions across datasets, and irregular geometry of detected footprints. LiDAR data, due to their increasing availability and geometric accuracy, offer significant potential for building footprint detection and updating. This paper aims to automatically or semi-automatically extract, vectorize, and regularize building footprints from classified point clouds, considering the accessibility of commonly used software tools. First, tools available in widely used GIS software environments (ArcGIS and QGIS) are applied, followed by the development of a custom Python-based program incorporating cluster analysis, polygonization, data cleaning, and regularization using available libraries and functions. All three approaches are applied in a case study (Bratislava, Slovakia) and evaluated by the aggregated shape similarity index using manually vectorized building footprints and cadastral data as an external data source. The results demonstrate that the proposed Python-based approach provides reliable and customizable outputs, particularly in contexts where only LiDAR data are available, while also supporting the integration of heterogeneous spatial data sources.
Ďuračiová et al. (Tue,) studied this question.