Automated Recognition of Pre-Cast Structural Elements from Point Cloud Data for Progress Monitoring Using BIM and LiDAR

Mehta, Ritik M.

Venkatachalam, Senthilkumar

2024-10-07T09:07:32Z

2024-10-07T09:07:32Z

2024

Automation techniques are increasingly being adopted in construction projects. However, the distributed nature of various automation and technological interventions in architecture, engineering, and construction (AEC) projects undermines their utility in various applications. Two such technologies, Building Information Modelling (BIM) and Light Detection and Ranging (LiDAR), have been demonstrated by various researchers, mostly independently. However, the potential synergy between these two technologies has been conceptually discussed in the literature and demonstrated in a few projects. Nevertheless, there are challenges in their wider application due to several reasons. One of the significant reasons is the level of automation that exists between these two technologies. The main focus is on the interoperability between the two technologies in accurately identifying various objects from the point-cloud data. The identification of construction objects plays a crucial role in the progress monitoring exercise. Progress monitoring is imperative to proactively identify deviations from the planned schedule for effective project management. As the demand for efficient progress monitoring grows, this study attempts to incorporate BIM with point cloud data generated from a terrestrial laser scanner to get weekly progress updates on a precast building construction. In this study, the developed framework superimposes the weekly clustered point cloud data (as-built) with the as-planned BIM model to recognize the 622 structural elements of a G+1 building (2000 sq m area), consisting mainly of precast structural walls and hollow core slabs for real-time object tracking. A total of 409 scans were performed to monitor the progress of precast structural elements over 13 weeks. Scanning locations were optimized based on the visual coverage of the progressed structural elements. Object recognition from the collected point cloud data was achieved through a suitable object recognition algorithm that superimposes the as-built point cloud with the as-planned BIM model. The framework achieved an overall precision of 93.26% and an overall recall value of 96.40% in recognizing the structural elements from the as-built point cloud data. Furthermore, the proposed framework uses earned value management principles to measure progress in terms of the volume of precast elements placed on-site against the baseline planned schedule, providing output in both visual and numerical formats.

angol

könyvfejezet

application/pdf

http://hdl.handle.net/10890/57730