Repository logo

Object Boundaries and Room Detection in As-Is BIM Models from Point Cloud Data

Accepted version


Conference Object

Change log


Anagnostopoulos, Ioannis 
Belsky, Michael 


Point CloudData (PCD)isextensively used for the creationof as-builtBuilding Information Modelsbecauseitaccurately depictsthe geometryof existing infrastructure. Modellers have to create this model from raw PCDmanuallyin the current practice.It isa laborintensive, time consuming,and error-prone processdue to the very high number of objects per model. In this paper,we tackle this issue through a novelsemi-automatedsolutionfor extracting object boundariesof as-ismodelsin Industry Foundation Classes (IFC) format from raw PCD. This method extracts the distinct boundaries of each object. Volumetric and parametric information are further minedfor spaces and objects respectively, such as such as volume, height and thickness of walls.Finally, the IFC model based on the pre-discoveredknowledgeis compiled.We tested this method on the creation of an IFC model for an office building. The results indicatethat an accurate IFC model can be achieved, however user input and adjustments are still necessary. Finally,semi-automation of the manual process can be achieved,leading todecreased laboreffort and cost.



Journal Title

Conference Name

Int. Conf. on Computing in Civil and Building Engineering

Journal ISSN

Volume Title


Publisher DOI

Publisher URL