Point Cloud Data Cleaning and Refining for 3D As-Built Modeling of Built Infrastructure

Abstract

Spatial sensing of built infrastructure is now a common practice within the AEC industry and results are commonly encapsulated in the form of dense point cloud data (PCD). PCD of built infrastructure might consist of millions of spatial points and it is well known that processing all these points is neither necessary nor computationally feasible. In addition, due to several reasons including hardware and/or software deficiencies, there might be several outliers that need to be removed from the PCD before further processing. As the result, cleaning and refining PCD is a paramount step in the process of spatial sensing and object-oriented modeling of built infrastructure scenes. This research work entails two parts: The first part provides an in-depth literature review on current states of practice and research on the concept of PCD cleaning. The second part presents the authors’ suggested framework for cleaning and refining PCD of built infrastructure. This prototype mainly consists of three major components: (1) removing outliers; (2) filling holes and gaps on surfaces of PCD; and (3) balancing the density of different areas of PCD based on a plane recognition approach. Several case studies are presented to demonstrate the efficiency of the proposed framework.