Computer vision-based CAD to BIM design change management: case of building's lightings

Abstract

Design changes are inevitable in construction projects and can arise for various reasons. When Building Information Modelling (BIM) is partially implemented during the design and construction, design changes are often introduced in traditional 2D plans. Consequently, these design changes must be incorporated into the corresponding BIM models before reissuing design drawings. Currently, a research gap exists in using advanced technologies, such as computer vision, alongside BIM for managing design changes. This study proposes a novel computer vision-based framework to automatically read and implement design changes from 2D plans into BIM models. Using real-world projects, a dataset containing 3742 images was generated to train three computer vision models to detect BIM elements and design changes. Accordingly, image alignment and text recognition techniques were deployed to read design changes and associate them with their respective BIM objects. A case study was implemented to validate the developed method. As a result, 92.9% accuracy was achieved in detecting the design changes from 2D plans and updating BIM models. This work contributes to the body of knowledge by introducing a novel method to automatically implement 2D design changes in BIM models and increasing the efficiency of the model updating process upon design change requests.