Automated Assessment of Reinforced Concrete Elements using Ground Penetrating Radar

Ground Penetrating Radar (GPR) has been evolving as a reliable Nondestructive tool for structural concrete inspections. Leveraging Electromagnetic waves enables the technique to be swift and advantageous for internal imaging of anomalies. With rebar and defect detections being the primary objective, post-processing the image/signal data for decluttered output is one of the major concerns. Availability of multiple GPR processing techniques on diverse applications make the appropriate technique selection a tough task. Traditionally structures were inspected for underlying defects and manually judged based on the semantic interpretation of radar signatures. However, cognitive decision making after processing enormous datasets can be time consuming and error prone. With advances in Computer Vision, there has been a surge in the number of neural architectures applied for automated object detection. This paper attempts to address the gray areas in technique optimization and automation by reviewing various GPR based manual detection models and their transition towards automated detection. Evolution of signal/image processing algorithms from manual migration-based imaging to automated object detection deploying Convolutional Neural Networks (CNNs) has been presented. This study also outlines various insights, challenges, and avenues for future research in the domain of non-invasive structural diagnostics using the GPR. • Comparison of multiple GPR processing techniques for detection of prevailing defects in concrete. • Reviewed various neural architectures for automatic detection of rebars and internal flaws in concrete. • Performance comparison of deep networks for rebars and defects detection in reinforced concrete. • Highlighted works on fusion of various NDE techniques with GPR. • Discussion on various factors affecting detectability of targets under the radar.