3DIR: exploiting topological relationships in three-dimensional information retrieval from BIM environments

摘要

An increasing amount of information is being packed into Building Information Models, with the 3D geometrical model serving as the central index leading to other building information. The Three-Dimensional Information Retrieval (3DIR) project investigates information retrieval from such environments, where information or documents are linked to a 3D artefact. Here, the 3D visualization/geometry can be exploited when formulating information retrieval queries, computing the relevance of information items to the query, or visualizing search results. Following reviews of literature in BIM and information retrieval, a clear gap was identified in the practice of information retrieval from BIM systems. The practical need for such a system was further specified using workshops with construction professionals as end users. A software prototype was developed, built on a commercial BIM platform. The 3DIR prototype creates an index of all text attached to the 3D model. The user can search for information by selecting specific 3D objects, specifying a spherical volume of the model and/or entering search keywords. This paper focuses on the exploitation of model topology. Relationships between 3D objects are used to widen the search, whereby relevant information items linked to a related 3D object (rather than information linked directly to a 3D object selected by the user) are still retrieved but ranked lower. Several such relationships between 3D objects were tested, whether explicitly encoded in the BIM information architecture or inferred from geometrical computations. An evaluation of the software prototype which exploits such topological relationships demonstrates its effectiveness but highlights the challenges to software users of added complexity. The system is subjectively rated comparably favorably. It is concluded that care needs to be taken when exploiting topological relationships, but that a tight coupling between text-based retrieval and the 3D model is generally effective in information retrieval from 3D BIM environments.