FINNMAP works with Aalto University to study the development of CityGML from 3D City Models
3D city models are increasingly an important instrument in numerous applications and solutions. Traditionally 3D city models have only represented the geometrical attributes of the given city, enabling easy visualisation. Yet, different thematic queries, analysis tasks, and spatial data mining are out of the reach of models that only offer geometrical information. CityGML 3D city models bring an addition of semantic information to the models.
Joonas Jokela of FINNMAP studied the process and different techniques of building reconstruction from airborne laser scanning as well as examined the process of developing CityGML from 3D building models. This study resulted into a thesis, which was supervised by Arttu Julin from FINNMAP, Doctor of Science Petri Rönnholm and Professor Henrik Haggrén from Aalto University. The main focus of this thesis was to study how well it is possible to automatically create CityGML 2.0 3D city models from data collected only by airborne laser scanning. The study was published at Aalto University on the 18th of January 2016.
CityGML has five different levels-of-detail indicating the level of precision of the building. LOD1 and LOD2 were the most important levels for this study, and hence it was tested how well different software were able to export reconstructed building models in the CityGML format with these precision levels. These exports were checked against the official specification of CityGML to see how well they met the requirements. In addition, the data and process requirements for creating higher precision level LOD3 was studied. Two different test areas were chosen with different building and roof types. One area included detached houses, some partly covered with vegetation, and another area included mainly apartment houses.
The Study clearly outlines the obstacles for automatically producing city models that are in line with the CityGML 2.0 standard. Joonas Jokela concludes that the model driven methods have problems when it cames to building installations, such as chimneys. As they are challenging to be modelled with software that uses model driven methods. Data driven methods on the other hand had problems when it came to the conversion from the building models to the CityGML format. Terrain and terrain intersection curve also turned out to be challenging. Additional software tool had to be developed to aid the conversion to CityGML.