Generating 3D City Models from Open LiDAR Point Clouds: Advancing Towards Smart City Applications

Abstract

In the past years, the amount of available open spatial data relevant to cities throughout the world has increased exponentially. Many cities, states, and countries have provided or are currently launching the provision of free and open geodata through public data portals, web-services, and APIs that are suitable for urban and smart cities applications. Besides ready to use 3D city models, many free and open LiDAR data sets are available. Several countries provide national LiDAR datasets of varying coverage and quality as free and open data. In this research, we introduce a novel pipeline to generate standardized CityGML conform Level of Detail (LoD)-2 city models for city-wide applications by using LiDAR generated point clouds and footprint polygons available from free and open data portals. Our method identifies the buildings and rooftop surfaces inside each footprint and classifies them into one of the five rooftop categories. When multiple buildings are present inside a footprint, it is divided into their corresponding zones using a novel corner-based outline generalization algorithm, addressing the need for more precise footprints and models in geometric and semantic terms. Finally, CityGML 2.0 models are created according to the selected category. This pipeline was tested and evaluated on a point cloud dataset which represent the urban area of the Spanish city of Logroño. The results show the effectiveness of the methodology in determining the rooftop category and the accuracy of the generated CityGML models.