Abstract:
Building Information Modeling (BIM) has benefited many value-added applications, including facility management, building retrofitting and renovation, among many others. However, the inevitable design changes, inadvertent deviations or errors, and renovation work make an as-designed BIM inappropriate for various applications. Thus, there is a need to reconstruct rich semantic as-built models. In recent years, generating rich semantic as-built BIMs has emerged as a significant research area but previous efforts failed to enrich BIM point clouds in real-time while using available color and texture information. Therefore, this research aims to advance the Derivative-Free Optimization-based (DFO) approach by automatically reconstructing rich semantic as-built BIM point clouds from a colored 3D point cloud. More specifically, this study focuses on evaluating the robustness of the proposed registration technique using (1) geometry and (2) geometry and color. Several experiments were conducted on indoor furniture, mechanical components, and precast concrete molds and the results revealed that the proposed method outperforms the previous methods in automatically generating rich semantic BIM point clouds
