A review on the integration of GPR and AI for enhanced road pavement inspection and precise crack detection

This review examines the integration of Ground Penetrating Radar (GPR) and Artificial Intelligence (AI) for non-destructive evaluation of road pavements. It focuses on high-frequency GPR antennas operating at 1 GHz or higher for subsurface imaging and on deep learning algorithms, including Convolutional Neural Networks (CNNs) and YOLO variants, for crack detection. Although crack detection is used as the main benchmark application, the review also addresses recurrent subsurface distresses in GPR-based pavement inspection workflows, including voids, interlayer debonding or delamination, and moisture-related damage. Recent developments in 3D modeling, pixel-level segmentation, amplitude-based analysis, and multimodal data fusion, such as integrating GPR with LiDAR or thermography, are evaluated in relation to performance, feasibility, and reliability. The review compiles validated case studies, identifies effective algorithmic strategies, and discusses measurable benefits, including improved detection precision, reduced manual effort, and better maintenance planning. It also examines implementation barriers, including data heterogeneity, environmental sensitivity, computational cost, and limited generalization of AI models across pavement types. Future research should prioritize standardized GPR datasets, including real, synthetic, and laboratory-based data, harmonized sensing protocols, and AI architectures capable of domain adaptation. Overall, GPR and AI integration provides a practical route toward proactive, data-driven pavement monitoring, resource optimization, and lifecycle management.