Skeleton-Based Registration of 3D Laser Scans for Automated Quality Assurance of Industrial Facilities

Registration of 3D point clouds is one possible way to compare the as-built and the as-designed status of construction components. Building information models (BIM) contain detailed information about the as-designed state, particularly 3D drawings of construction components. On the other hand, using automated and accurate data acquisition methods such as laser scanning provide reliable and robust information about the as-built status of construction components. Registration therefore makes it possible to automatically compare the designed and built states in order to appropriately plan forward and generate the corrective actions required. This paper presents a new approach for reliably performing the registration with a required level of accuracy and automation within a substantially improved timeframe. Rather than performing the computationally intensive registration methods that may not work robustly for dense point clouds, the proposed framework employs the geometric skeleton of the construction components, which is extremely less dense and therefore computationally less costly for the processing step required. The method is experimentally tested for the components extruded along an axis, such as industrial assemblies (i.e. pipe spools and structural frames) for which a geometric skeleton represents the components abstractly. The registration of 3D point clouds is performed in a computationally less intensive manner, and the framework developed has the potential to be employed for (near) real-time assembly control, quality control and status assessment processes.