<p>Hidden voids are a type of subgrade distress in urban roads, characterized by their high concealment and detrimental nature. Therefore, research on GPR (Ground Penetrating Radar) forward modeling and intelligent identification for subgrade distress characterized by hidden voids has become one of the most prominent and challenging problems in the field. This study focuses on the GPR forward modeling and intelligent identification methods for subgrade distress characterized by hidden voids. Firstly, the Finite-Difference Time-Domain (FDTD) method was employed to analyze the GPR image features of subgrade distress under various conditions. These conditions include hidden voids, irregular hidden voids (air-filled and water-filled), hidden voids coupled with brick-concrete manholes, and hidden voids coupled with PE or reinforced concrete (RC) pipe culverts, as well as their corresponding relationships with void parameters. Subsequently, based on field GPR detection and forward modeling results, an image database of subgrade distress was established. Following this, detection experiments for hidden voids in subgrade distress images were conducted using three object detection networks: YOLO v3, YOLO v5, and YOLO v8. Finally, the training iteration patterns of the Loss function and mean Average Precision (mAP), along with their prediction accuracy, were analyzed. The prediction accuracy of all three object detection algorithms for hidden void distress exceeded 0.71, successfully achieving the research objective of intelligent identification for this type of subgrade distress. To address the issue of hidden void distress in subgrade, this study integrates theoretical analysis, forward modeling, field detection, and the investigation of recognition algorithms. Through this comprehensive research framework, the scientific validity, rationality, and practical applicability of the proposed intelligent identification method for hidden void distress have been successfully demonstrated.</p>