Automated Site Extraction for Scan-to-BIM Progress Monitoring in Road Network Construction

Abstract

Construction progress monitoring is an essential activity for ensuring a project’s quality and timely delivery. Scan-to-BIM demonstrates its potential and superiority in capturing the as-built status and handling design changes to achieve real-time project tracking. Previous studies have focused on its main steps, including segmentation, recognition, and reconstruction, but often overlooked target area extraction due to their emphasis on residential buildings. Manual extraction can be tedious, time-consuming, and labor-intensive, especially for large-scale, horizontal transport infrastructures. Although similar applications can be found in other domains, these methods may not be feasible for extracting road networks construction sites due to differences in data structure, the as-built status of roads, and the requirements for the Level of Detail (LoD) in the output. This paper proposes an automated hybrid road construction site extraction framework as the input for subsequent Scan-to-BIM construction progress monitoring. It leverages both local feature analysis and graph theory. The developed road patch generation and connection processes enable the extraction of complex road networks with high generality and versatility. Two sets of Airborne Laser Scanning (ALS) data with different construction progress were used to validate the proposed framework, achieving site extraction accuracies of 99.59% and 94.50%, respectively. These results indicate that the proposed method can effectively serve as a basis for supporting construction progress monitoring through Scan-to-BIM.