Community Research and Development Information Service - CORDIS

FP7

BIMAUTOGEN Result In Brief

Project ID: 247586
Funded under: FP7-PEOPLE
Country: United Kingdom

New research takes building modelling to the next level

A consortium of partners from different fields and continents has helped launch new architecture and infrastructure software for Building Information Modelling.
New research takes building modelling to the next level
Building Information Modelling (BIM), the digital representation of physical and functional characteristics of a facility, offers valuable support for making decisions about buildings during their life cycle. From initial construction and design to retrofitting installations and recycling construction material, BIM offers powerful 3D visualisation that also considers time and cost involved. Today, few buildings and construction projects have comprehensive digital records of how they were built, prompting a need to create BIM software for existing and future construction.

The EU-funded BIMAUTOGEN (Collaboration for Research and Education in Automated Generation of Building Information Models) project aimed to develop new software for scanning buildings using high-tech BIM. It brought together experts in scanning, videogrammetry, computer imaging, machine learning, and parametric object modelling to automatically create ‘as-is modelling’ for all kinds of buildings.

To achieve its technical objectives the team adopted a new approach to generate infrastructure point cloud data using videogrammetry, integrating novel data cleaning methods that regulate point density and address data holes and outliers. It then outlined a machine learning approach for extracting objects from the point cloud data and features, focusing on bridges, buildings and industrial plants.

The latter phase of the software development involved a new method to fit BIM object templates into the extracted high level primitives. It also involved validating initial object identification related to the physical juxtaposition of the objects in question. Lastly, the team built a semantic relationship engine to detect the type of connections between objects based on predefined rules.

These efforts led to a new videogrammetric framework which was released on the market to generate highly accurate building measurements. The newly developed toolset for extracting high-level primitives (planes, cylinders, cones etc.) from point cloud datasets has also been adopted by users. Plans to integrate the different software components by the project’s partners beyond the end of the initiative were then established.

The project’s results have been published in several journals and disseminated to relevant stakeholders, and have already made an impact in upgrading as-is modelling research. BIM will no doubt strengthen today’s architecture practices and improve infrastructure life cycle management once its full potential is realised.

Related information

Keywords

Building Information Modelling, BIM, videogrammetry, computer imaging, machine learning
Record Number: 182918 / Last updated on: 2016-06-23
Domain: Industrial Technologies