Skip to main content
European Commission logo print header

Collaboration for Research and Education in Automated Generation of Building Information Models

Final Report Summary - BIMAUTOGEN (Collaboration for Research and Education in Automated Generation of Building Information Models)

Only very few constructed facilities today have a complete record of as-built information. Despite the growing use of Building Information Modelling and the improvement in as-built records, several more years will be required before guidelines that require as-built data modelling will be implemented for the majority of constructed facilities, and this will still not address the stock of existing buildings. A technical solution for scanning buildings and compiling Building Information Models is needed. However, this is a multidisciplinary problem, requiring expertise in scanning, computer vision and Videogrammetry, machine learning, and parametric object modelling. The research objective of the BIMAUTOGEN project has been the facilitation of international collaboration and transfer of knowledge between international partners and the validation of the hypothesis that the project's proposed novel framework presented can be successfully used to generate parametric building models of buildings, ranging from residential housing to industrial facilities, almost entirely automatically.

The work that has been performed since the beginning of the project spans across all seven work packages. In WP1 the UCAM team, in collaboration with FORTH and Georgia Tech created a novel method for generating infrastructure point cloud data from either monocular or stereo-Videogrammetry. The new method takes into account line and plane features that are much more robust in infrastructure scenes to improve the optimization results of bundle adjustment. In WP2 the team above enhanced the quality of the point cloud data generated with novel data cleaning methods that regulate point density and address data holes and outliers. In WP3 the UCAM team created a novel framework for infrastructure features detection that enabled the creation of various pattern recognition models for both infrastructure objects and defects. In WP4, the Georgia Tech team, in collaboration with UCAM and UCY, created a machine learning approach to extracting objects from the point cloud data (WP2) and features (WP3). This approach was focused on bridges as a first case study, and has instigated similar research on buildings and industrial plants. In WP5, UCAM and Stanford collaborated to test a novel approach to fit BIM object templates into the extracted high level primitives of WP4. In WP6, work was performed by Technion in collaboration with Georgia Tech and UCY to develop a system that applies a set of topological rules to check whether the initial object identification is reasonable in terms of the physical juxtaposition of the objects. This was coupled by the work in WP7 by the same team that led to a semantic relationship engine that can detect the type of connections between objects based on a predefined ruleset.


The proposed work is now complete, and the final results of all WPs have been published or submitted for publication in several journals and conferences. The last few remaining publications are on schedule to be submitted for review within 2 months from the project’s completion. Specifically, WPs 1 and 2 have led to an already commercialized Videogrammetric framework that is able to generate building measurements at fabrication accuracy. This is now available to the public through start-up company Pointivo. WPs 3-5 have created a basic toolset for detecting high level primitives (i.e. planes, cylinders, cones, etc.) from point cloud datasets. This toolset has been widely published and is now being used a benchmark for comparison with new tools that are incrementally improving the original set. WP6-7 have created a modelling toolset that can import high level primitives and convert them into a useful geometric BIM. The initially proposed work package plan did not include an integration work package that would attempt to integrate all the prototypes developed by the consortium subgroups, hence the individual parts are not integrated yet. However, this cross-continent collaboration led to additional proposal writing which led to securing an EC Infravation grant (“SeeBridge”) mostly by members of this consortium. The scope of the new project is to integrate the parts and create a single, robust tool for as-is modelling in the case of bridges.

Overall, this project has been instrumental in setting the foundation tools for as-is modelling research, and has instigated multiple follow up efforts by sub-groups of the consortium members as well as other researchers. It is fair to say that this project put “as-is modelling” research on the map. This has also occurred in a very timely manner; as-is models are needed for BIM to realise its full potential particularly in the operation and maintenance phase of the infrastructure life-cycle. The project’s most significant contributions have been, not only the automation of several mundane and repetitive processes with the addition of visual and spatial pattern recognition concepts in the modelling workflow, but also the exchange of knowledge and the building of transatlantic research collaborations on cutting-edge and high-impact scientific projects through the exchange of interdisciplinary staff among partner institutions and joint training of research teams on thematic areas of common research interest.