Cost Effective Rebar Inspection using Point Cloud Data

Traditional methods for quality inspection of reinforcing bars (rebars) in reinforced concrete (RC) structures are labor-intensive, time-consuming, and pose safety risks. Recent methods utilizing advanced sensor technologies, despite their effectiveness, are often too costly, limiting their widespread use. Additionally, vision-based alternatives often sacrifice accuracy in measurements and geometry. In response to these challenges, this paper introduces a robust framework that uses point cloud data captured by iPhone LiDAR sensor for the efficient and cost-effective rebar inspection. Our approach is articulated through two principal modules: the Component Recognition Module (CRM) and the Dimensional Assessment Module (DAM). The CRM incorporates an adapted RandLA-Net model for the semantic segmentation of point clouds, enabling the precise identification of rebar components. Meanwhile, the DAM employs an adapted RANSAC algorithm for evaluating rebar spacing. Based on our framework's evaluation, we achieved an overall accuracy of 82.1% in the semantic segmentation task and a mean absolute error of 0.9 cm in the rebar spacing evaluation task. The results demonstrate that our proposed framework provides accurate rebar inspection while remaining cost-effective.