Preliminary Study on Automating Rebar Detection in Reinforced Concrete Structures Using YOLOv11 and GPR Data

Ground-penetrating radar (GPR) is a widely used non-destructive testing (NDT) technique for concrete inspections and locating reinforcement bars (rebars). However, manually detecting rebars in GPR data is labor-intensive and timeconsuming, limiting the efficiency of large-scale structural evaluations. Recent advances in deep learning, particularly object detection algorithms, provide a solution to automate this process, enabling engineers to focus on further analysis. This study explores the capabilities of YOLOv11, an advanced object detection model, to automate rebar detection in GPR data. The model was trained on diverse public datasets to ensure robustness. Performance evaluations show that YOLOv11-x achieves a mean average precision (mAP) of 99.40%} at 50% IoU (mAP50) and 86.20% across multiple IoU thresholds (mAP50-90), with a precision of 90.04 % and recall of 98.99 %. The model was validated using an independent dataset from a GPR survey of a reinforced concrete pedestrian bridge, demonstrating its potential to improve both the accuracy and scalability of rebar detection. YOLOv11 was also compared with YOLO v8, v9, v10, and v12, highlighting its improvements in precision, efficiency, and generalization across GPR datasets.