Road Design Layer Detection in Point Cloud Data for Construction Progress Monitoring

Abstract

Poor performance in transportation construction is well documented, with an estimated $114.3 billion in global annual cost overrun. Studies aimed at identifying the causes highlighted traditional project management functions like progress monitoring as the most important contributing factors. Current methods for monitoring progress on road construction sites are not accurate, consistent, reliable, or timely enough to enable effective project control decisions. Automating this process can address these inefficiencies. The detection of layered design surfaces in digital as-built data is an essential step in this automation. A number of recent studies, mostly focused on structural building elements, aimed to accomplish similar detection but the methods proposed are either ill suited for transportation projects or require labeled as-built data that can be costly and time consuming to produce. This paper proposes and experimentally validates a model-guided hierarchical space partitioning data structure for accomplishing this detection in discrete regions of three-dimensional as-built data. The proposed solution achieved an F1 score of 95.2% on real-world data, confirming the suitability of this approach.