Ground Penetrating Radar for Subsurface Utility Detection: Methods, Challenges, and Future Directions.

Area of Science:

• Geophysics
• Urban Infrastructure Management
• Artificial Intelligence

Background:

• Ground-penetrating radar (GPR) is vital for subsurface utility mapping in urban environments.
• Accurate detection of buried pipelines and cables is crucial for excavation safety and infrastructure management.
• Existing GPR methods face challenges like event-utility mismatch and domain gap between synthetic and real-world data.

Purpose of the Study:

• To review the state-of-the-art in GPR for urban subsurface utility mapping.
• To identify key barriers hindering accurate utility detection and inference.
• To propose future research directions for improving GPR reliability in urban settings.

Main Methods:

• Bibliometric analysis of deep learning applications in GPR.
• Review of challenges in urban GPR data interpretation, including clutter and signal variability.
• Analysis of limitations in synthetic data generalization for real-world field conditions.

Main Results:

• Deep learning is increasingly used, but primarily for event detection, not utility-level inference.
• Real urban GPR data presents complexities: orientation-dependent signatures, clutter, and non-utility anomalies.
• Synthetic data often fails to generalize due to unrepresented soil heterogeneity and acquisition variability.

Conclusions:

• Future progress requires shifting from event detection to utility-level reasoning.
• Multi-sensor fusion, physics-guided learning, and hybrid datasets are crucial.
• Uncertainty-aware interpretation is essential for actionable GPR outputs in urban engineering.