Skip to main content

LUP Student Papers

LUND UNIVERSITY LIBRARIES

Generating 3D building models according to Swedish building specification using footprints and airborne laser scanning

Klasson, Andreas LU (2025) In Master Thesis in Geographical Information Science GISM01 20251
Dept of Physical Geography and Ecosystem Science
Abstract
Buildings are a central component of three-dimensional (3D) city models. Generating such models according to an appropriate standard is not always straightforward and can be time-consuming. This study explores a method for generating building models that comply with the Swedish national specification (NS) for buildings, at a level of detail (LOD) that includes roof structures. The method evaluates how building footprints from the national building register, combined with open airborne laser scanning (ALS) data with a point density of 3.7 points/m2, can be used to generate buildings with flat and gable roofs. A semi-automatic approach is used, applying the RANSAC algorithm to detect roof planes and calculate their intersections.
The... (More)
Buildings are a central component of three-dimensional (3D) city models. Generating such models according to an appropriate standard is not always straightforward and can be time-consuming. This study explores a method for generating building models that comply with the Swedish national specification (NS) for buildings, at a level of detail (LOD) that includes roof structures. The method evaluates how building footprints from the national building register, combined with open airborne laser scanning (ALS) data with a point density of 3.7 points/m2, can be used to generate buildings with flat and gable roofs. A semi-automatic approach is used, applying the RANSAC algorithm to detect roof planes and calculate their intersections.
The results show that the successfully generated buildings had a root mean square error (RMSE) in height of 0.15 m for gable roofs and 0.21 m for flat roofs, when compared to reference points. For gable roofs, the ridge endpoint location had a planar RMSE of 0.09 m. However, only simple roof shapes and structures could be accurately generated. Buildings with complex roof geometries or varying roof heights were not successfully modelled. The planar accuracy reflects the high quality of the building footprint data, while the height accuracy demonstrates the method’s suitability for simple gable and flat roofs on buildings that do not require further segmentation due to height differences. This study highlights a viable approach for generating accurate building models of simple structures that comply with the Swedish NS building standard. (Less)
Abstract (Swedish)
Byggnader är en central komponent i 3D stadsmodeller. Att skapa sådana modeller enligt relevant standard är inte alltid enkelt och kan vara tidskrävande. Den här studien undersöker en metod att skapa byggnadsmodeller utifrån the svenska nationella specifikationen (NS) för byggnader, i en detaljnivå (LOD) som inkluderar takform. Metoden utvärderar hur byggnadsytor (registerbyggnad) tillsammans med flygburen laserskannings data, med punkttäthet 3.7 punkter/m2, kan generera byggnadsmodeller med platt eller sadeltak. En semi-automatiskt metod används där RANSAC-algoritmen tillämpas för att identifiera takplan och beräkna skärningslinjer.
Resultatet visar att de byggnader som lyckades att genereras korrekt hade ett rotmedelkvadratfel (RMSE) i... (More)
Byggnader är en central komponent i 3D stadsmodeller. Att skapa sådana modeller enligt relevant standard är inte alltid enkelt och kan vara tidskrävande. Den här studien undersöker en metod att skapa byggnadsmodeller utifrån the svenska nationella specifikationen (NS) för byggnader, i en detaljnivå (LOD) som inkluderar takform. Metoden utvärderar hur byggnadsytor (registerbyggnad) tillsammans med flygburen laserskannings data, med punkttäthet 3.7 punkter/m2, kan generera byggnadsmodeller med platt eller sadeltak. En semi-automatiskt metod används där RANSAC-algoritmen tillämpas för att identifiera takplan och beräkna skärningslinjer.
Resultatet visar att de byggnader som lyckades att genereras korrekt hade ett rotmedelkvadratfel (RMSE) i höjd på 0,15 m för sadeltak och 0,21 m för platta tak, vid jämförelse mot referenspunkter. Sadeltakens nock-ändpunkter hade ett RMSE på 0,09 m i plan. Däremot kunde endast enkla takformer och byggnadskonstruktioner genereras. Byggnader med komplexa takgeometrier eller varierande takhöjder kunde inte genereras. Det låga RMSE-värdet i plan återspeglar den höga kvaliteten på byggnadsytorna, medan RMSE i höjd visar metodens lämplighet för enklare sadeltak eller platta tak som inte kräver ytterligare segmentering på grund av höjdskillnader. Studien visar en fungerande metod för att generera noggranna byggnadsmodeller av enklare takstruktur som följer den svenska standarden NS byggnad. (Less)
Popular Abstract
Digital 3D models of cities are becoming more important in many areas, including urban planning, environmental studies, and infrastructure management. These models can contain many different things, such as roads, trees, and terrain, but buildings usually play the most central role. To create detailed and realistic building models, especially ones that include roof shapes, we need good data and a clear structure for how the models should be built.

In Sweden, there is a national standard that describes how digital building models should be created and what information they should include. This helps ensure that the models can be used and shared more easily between different organisations. However, creating models that follow this... (More)
Digital 3D models of cities are becoming more important in many areas, including urban planning, environmental studies, and infrastructure management. These models can contain many different things, such as roads, trees, and terrain, but buildings usually play the most central role. To create detailed and realistic building models, especially ones that include roof shapes, we need good data and a clear structure for how the models should be built.

In Sweden, there is a national standard that describes how digital building models should be created and what information they should include. This helps ensure that the models can be used and shared more easily between different organisations. However, creating models that follow this standard can be difficult and time-consuming, especially when it comes to modelling roofs, since roof details are not included in standard 2D maps of which building areas usually is shown.
In my thesis, I explored a way to automatically generate 3D building models that follow the Swedish standard. I used two types of open data: 2D building area footprints from the national building register and airborne laser scanning data that provides information about the shape and height of buildings. I focused on simple roof types, such as flat and gable roofs, which are common in Sweden. A algorithm was used to detect the shape of the roofs and then create 3D models from that information together with the footprint.
The method gave good results for buildings with simpler roof shapes. The models were accurate in both height and planar position when compared with reference measurements. However, the method did not work as well for more complex roofs that had several levels or unusual shapes. These more complex roof shapes were considered unsuccessful because the roof geometry could not be created properly.
This project shows that it is possible to create reliable 3D models of many buildings in Sweden using only open and freely available data. This could make it much easier and faster to build 3D city models, which can be useful for a wide range of applications in the future. (Less)
Please use this url to cite or link to this publication:
author
Klasson, Andreas LU
supervisor
organization
course
GISM01 20251
year
type
H2 - Master's Degree (Two Years)
subject
keywords
Geography, GIS, Arial Laser Scanning, ALS, Building model, Level of Detail, LOD, RANSAC, National Specification Building
publication/series
Master Thesis in Geographical Information Science
report number
197
language
English
id
9213981
date added to LUP
2025-10-16 16:16:57
date last changed
2025-10-16 16:16:57
@misc{9213981,
  abstract     = {{Buildings are a central component of three-dimensional (3D) city models. Generating such models according to an appropriate standard is not always straightforward and can be time-consuming. This study explores a method for generating building models that comply with the Swedish national specification (NS) for buildings, at a level of detail (LOD) that includes roof structures. The method evaluates how building footprints from the national building register, combined with open airborne laser scanning (ALS) data with a point density of 3.7 points/m2, can be used to generate buildings with flat and gable roofs. A semi-automatic approach is used, applying the RANSAC algorithm to detect roof planes and calculate their intersections.
The results show that the successfully generated buildings had a root mean square error (RMSE) in height of 0.15 m for gable roofs and 0.21 m for flat roofs, when compared to reference points. For gable roofs, the ridge endpoint location had a planar RMSE of 0.09 m. However, only simple roof shapes and structures could be accurately generated. Buildings with complex roof geometries or varying roof heights were not successfully modelled. The planar accuracy reflects the high quality of the building footprint data, while the height accuracy demonstrates the method’s suitability for simple gable and flat roofs on buildings that do not require further segmentation due to height differences. This study highlights a viable approach for generating accurate building models of simple structures that comply with the Swedish NS building standard.}},
  author       = {{Klasson, Andreas}},
  language     = {{eng}},
  note         = {{Student Paper}},
  series       = {{Master Thesis in Geographical Information Science}},
  title        = {{Generating 3D building models according to Swedish building specification using footprints and airborne laser scanning}},
  year         = {{2025}},
}