Automatic LOD2 3D Building Modelling for Historical Cities

This repository contains code and scripts developed for the study on automated generation of LOD2 3D building models, demonstrated in the historical city of Olomouc (Czechia). The project integrates LiDAR data, orthoimagery, and building footprints to automate key steps such as roof type classification, building height estimation, and procedural modelling of complex roof structures. The outputs support applications in 3D geovisualization, urban planning, cultural heritage documentation, and spatial analysis.

Dataset description:
Height_assessment.ipynb – Python notebook for assessing the accuracy of ridge and eave height estimation, including RMSE calculations and outlier analysis.
Roof_assessment.ipynb – Python notebook for evaluating and determining optimal thresholds for roof classification based on slope indicators.
Building_test3.cga – Computer-Generated Architecture (CGA) script for Esri CityEngine used to create procedural rules for LOD2 building generation, supporting multiple roof types and customizable parameters.
Parameters.png - Parametres of CGA script

The model can be viewed here: https://bit.ly/3D_scenes
Esri ArcGIS Pro project is available here: http://eyetracking.upol.cz/3Droofs-ProjectFiles.zip