AI-based recognition for road distressed from GPR measurement using artificial neural networks

Gprmax is utilized for ground-penetrating radar simulation, enabling to generate imagery delineating diverse road pathologies. The autonomous identification and classification of these road afflictions are performed leveraging a dataset trained within the YOLO framework. Renowned for its efficacy in target detection, YOLO harnesses pre-trained model parameters from extensive databases, thereby achieving commendable recognition outcomes. To ameliorate the protracted training duration, the adoption of the random retreat mechanism is advocated. The methodology elucidated in this investigation attains an average recognition rate of 77.83%, effectively achieving the identification of road disease images and thereby enhancing operational efficiency. Significantly, the proposed approach exhibits promising prospects for application and dissemination within engineering domains.