Development of Feature Extraction Using High-Resolution RGB Images and Homography Transformation to Inspect RC Walls Rebar Works

초록

Reinforcing bars (rebars) are essential for resisting tensile forces and ensuring structural safety in reinforced concrete systems. However, current rebar inspection practices rely on manual and sampling-based measurements, leading to human errors and limiting inspection coverage. Although automated 3D and LiDAR-based inspection systems have been explored, their adoption rate remains low due to high cost and workflow complexity. To address these challenges, this study presents a cost-efficient vision-based inspection framework that combines high-resolution image analysis and marker-assisted geometric correction. The proposed method detects rebar and marker regions in low-resolution images, refines intersection keypoints in high resolution, performs marker-based rectification, and computes rebar spacing, diameter, and quantity from the corrected geometry. Experiments under varied distances and viewing angles demonstrated accurate and reliable performance, achieving 100% accuracy in counting vertical and horizontal rebars, mean absolute errors of 0.59 mm for diameter and 4.05 mm for spacing. These results confirm practical, low-cost automated rebar measurements for construction inspection.

키워드