Automatic Road Subsurface Distress Recognition from Ground Penetrating Radar Images using Deep Learning-based Cross-verification

2025-07-15

AI Summary

Road subsurface defects (RSDs) are conventionally identified manually, resulting in low efficiency and poor generalizability. Method: This paper proposes a deep learning–based automated detection framework leveraging multi-view 3D ground-penetrating radar (GPR) images. We introduce a large-scale, field-collected, and expert-annotated 3D GPR dataset and design a YOLO-driven cross-view feature validation mechanism that exploits complementary sensitivity of distinct scanning perspectives to defect morphology. Contribution/Results: The proposed method achieves a recall rate exceeding 98.6% under complex field conditions, significantly enhancing detection robustness and completeness. Upon online deployment, it reduces manual inspection effort by approximately 90%. By integrating multi-view geometric reasoning with scalable deep learning, our approach establishes a transferable technical paradigm for intelligent interpretation of GPR imagery.

Technology Category

• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Machine Learning: Multi-instance/Multi-view Learning
• Computer Vision: Vision for Robotics & Autonomous Driving

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Ground penetrating radar (GPR) has become a rapid and non-destructive solution for road subsurface distress (RSD) detection. However, RSD recognition from GPR images is labor-intensive and heavily relies on inspectors' expertise. Deep learning offers the possibility for automatic RSD recognition, but its current performance is limited by two factors: Scarcity of high-quality dataset for network training and insufficient capability of network to distinguish RSD. In this study, a rigorously validated 3D GPR dataset containing 2134 samples of diverse types was constructed through field scanning. Based on the finding that the YOLO model trained with one of the three scans of GPR images exhibits varying sensitivity to specific type of RSD, we proposed a novel cross-verification strategy with outstanding accuracy in RSD recognition, achieving recall over 98.6% in field tests. The approach, integrated into an online RSD detection system, can reduce the labor of inspection by around 90%.

Problem

• Automating road subsurface distress recognition from GPR images
• Overcoming scarcity of high-quality GPR training datasets
• Enhancing deep learning accuracy for RSD detection

Innovation

• Deep learning-based cross-verification for RSD recognition
• Rigorous 3D GPR dataset with 2134 diverse samples
• Online RSD detection system reducing labor by 90%