Abstract
Automated progress monitoring builds an important foundation for objective productivity analysis of construction processes. Digital twins of the construction phase rely on fully automated approaches to acquire near real-time progress information. This is essential for identifying bottlenecks during construction and supporting future project planning. Many existing vision-based methods lack automated image acquisition, fast computation times, or fine-grained progress information. This paper presents a new vision-based construction monitoring approach that reduces the geometric information provided in exchange for a higher time resolution and a higher level of automation. Instead of the detailed geometry, the real-time status of the building elements is provided. It is applied to cast-in-place concrete columns, identifying individual operational steps. The approach is based on projecting building elements from a building model onto images of a fixed on-site camera to then classify them according to the current element status with the help of a CNN. Using image sequences additionally allows accounting for moving objects and other outliers, which makes the approach robust and reliable.
Original language | English |
---|---|
State | Published - 2023 |
Event | 30th International Conference on Intelligent Computing in Engineering 2023, EG-ICE 2023 - London, United Kingdom Duration: 4 Jul 2023 → 7 Jul 2023 |
Conference
Conference | 30th International Conference on Intelligent Computing in Engineering 2023, EG-ICE 2023 |
---|---|
Country/Territory | United Kingdom |
City | London |
Period | 4/07/23 → 7/07/23 |