Semantic Enhanced As-Built BIM Updating Based on vSLAM and Image Processing

Building information model (BIM) needs to be updated based on the real situation on the construction site to reflect the dynamic aspects of construction progress. However, without realizing real-time tracking technology and semantic enhancement for the 3D reconstructed model, it is difficult to achieve an updating based on the comparison of the as-designed BIM with the as-built BIM for any discrepancy in terms of quality and quantity. Moreover, semantic information is usually missing in the re-constructed models; therefore, it is difficult to automate the process of identifying target building components in a complicated construction site. The present paper proposes an image-based 3D reconstruction while integrating infrared thermography to extract the semantic information from the images as the different emissivity of the construction materials is a piece of crucial evidence for the identification of the material. Case studies are conducted in an under passage in a university campus to investigate the feasibility of the proposed methods. An efficient and cost-benefit approach of as-built BIM updating is attainable considering the future possibility of high-level semantic information reasoning from different data sources. The results of the proposed methodology show that the thermal images perform better in capturing the MEP data comparing to the 3D reconstruction using images where MEP data are missing and incomplete in the point cloud. However, the limitations of the methodology, such as the inconsistent thermal conditions and the unpredictable thermal radiation source will influence the accuracy of segmentation results.