Road Layer Detection and Volume Calculation Using UAV Technologies and Artificial Intelligence

Abstract

Traditional methods for monitoring road construction often rely on manual measurements and visual inspections, which are time-consuming and prone to errors. This study introduces a novel approach that uses unmanned aerial vehicles (UAVs) and the You Only Look Once (YOLO) model (YOLOv11-seg) to significantly improve the accuracy and efficiency of road layer detection and volume calculation. By utilizing high-resolution images taken by the DJI Phantom 4 Pro drone, the model successfully identifies and segments crucial road layers, such as subgrade, asphalt, and aggregate base. The YOLOv11-seg model demonstrates remarkable classification accuracy, achieving 95% for both asphalt and aggregate base and 83% for subgrade. Additionally, the model reaches a maximum precision of 1.0 at the highest confidence level. Its F1 score is 0.92 at an optimal confidence threshold of 0.779, and it maintains a recall of 0.96 at lower confidence levels. For volumetric analysis, this study evaluates four software tools: Agisoft Metashape, CloudCompare, Civil 3D, and WebODM. It identifies WebODM as the most accurate, achieving volume accuracies of 99.85% for subgrade and asphalt and 99.9% for aggregate base. This application of drone imagery and object detection technology simplifies monitoring and significantly reduces labor and errors associated with traditional road construction monitoring methods.