Semi-Automatic Pipe Network Reconstruction Using Point Cloud Data

As the use of Building Information Modeling (BIM) for retrofits and Operations and Maintenance (O&M) activities in existing buildings becomes increasingly widespread, modeling the as-is building conditions becomes a requirement. Pipes are among the elements of most interest in these endeavors because they represent a significant portion of the building systems and have major impacts on O&M efforts and expenditures. Given the usually complex configurations of piping systems in existing commercial and industrial facilities and the tedious, time-consuming, and error-prone process associated with manual BIM modeling, automated methods for pipe modeling from point clouds have been proposed, which drastically reduced modeling times. A significant limitation that persists, however, refers to the fact that the classification of the piping systems is still a manual process in all current applications. While other researchers have reconstructed pipes, none have provided an automated method to capture and include the critical semantic information pertaining to the system type and usage. This paper, which is the first part of a major project that aims to reconstruct and classify piping systems in existing buildings using data from laser scanners and thermal cameras, shows the initial results of the proposed method for piping system reconstruction. The proposed reconstruction method is based on the estimation of the skeletons of straight pipes, followed by the identification of tees and elbows based on the relationships among straight pipe segments. The main results include the computation times for the process and the comparison of the as-is model to the reconstructed BIM model.