Semantic enrichment is a process of supplementing/correcting information in a poorly prepared BIM model. Object classifications are essential information, but are commonly missing or incorrectly represented when transferring a BIM model or creating a model using tools customized for other domains in design. Automated compilation of 'as-is' BIM models from point cloud data also requires object classification, as well as 3D reconstruction. We present a systematic approach to classifying objects in a BIM model, for use in future semantic enrichment systems. Previous work on object classification in BIM model enrichment was restricted by its limited ability to accurately interpret geometric and spatial features and by the constraints of Boolean logic rules and the rule compilation process. To address these issues, we propose a procedure for establishing a knowledge base that associates objects with their features and relationships, and a matching algorithm based on a similarity measurement between the knowledge base and facts. An implementation on a synthetic bridge model shows that whereas some objects can be classified by shape features alone, most objects require the use of spatial relations for unique classification. Spatial context is more likely to uniquely identify an object than shape features are.
|Title of host publication||Proceedings of the 2017 Joint Conference on Computing in Construction (JC3)|
|Editors||Frédéric Bosché, Ioannis Brilakis, Rafael Sacks|
|Place of Publication||Edinburgh|
|Number of pages||8|
|Publication status||Published - 2017|
Ma, L., Sacks, R., & Kattel, U. (2017). Building Model Object Classification for Semantic Enrichment Using Geometric Features and Pairwise Spatial Relations. In F. Bosché, I. Brilakis, & R. Sacks (Eds.), Proceedings of the 2017 Joint Conference on Computing in Construction (JC3) (Vol. 1, pp. 373-380). Heriot-Watt University. http://itc.scix.net/data/works/att/LC3_2017_paper_044.pdf