Automated earthwork progress monitoring for construction projects

Date

2023-8-10

Author

Ersöz, Ahmet Bahaddin

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

274
views

158
downloads

Cite This

This thesis presents a novel approach to earthwork progress monitoring on construction sites using Unmanned Aerial Vehicles (UAVs), machine learning, and computer vision. Traditional methods, often time-consuming and prone to error, are addressed through the developed methodology, enhancing efficiency and accuracy. A fully autonomous pipeline is established, from image acquisition to volume calculation. The utilization of UAVs for imaging during operational hours improves safety measures by eliminating the risks associated with manual inspections. A real-world aerial image dataset with pixel-level annotation from a wide variety of construction environments annotation is presented. A machine learning model is trained using this dataset that isolates construction machinery from images, achieving an average precision score exceeding 90% for typical machinery categories. An image inpainting algorithm is presented to remove these machines from images, enhancing the visual representation of construction terrains for volume calculation. Furthermore, a novel volume calculation algorithm is proposed, providing an accurate and efficient method for estimating earthwork volumes. This study performs a comparative experiment of point clouds generated by UAVs and Terrestrial Laser Scanning, demonstrating the UAV-based approach's viability as a reliable and cost-effective alternative. Despite these advancements, the research acknowledges challenges in UAV performance due to weather variations, difficulties detecting diverse construction machinery, and the complexity of dealing with closely positioned machines. In conclusion, this research paves the way for more efficient construction management practices. Future research can focus on expanding the machine learning training datasets to improve the process further and integration with 4D Building Information Modeling (BIM).

Subject Keywords

Earthwork Progress Tracking, Construction Image Dataset, Unmanned Aerial Vehicles, Deep Learning, Image inpainting

URI

https://hdl.handle.net/11511/104892

Collections

Graduate School of Natural and Applied Sciences, Thesis