Abstract
Urban roads and infrastructure, vital to city operations, face growing threats from subsurface anomalies like cracks and cavities. Ground Penetrating Radar effectively visualizes underground conditions using electromagnetic waves; however, accurate anomaly detection through this method remains challenging due to limited labeled data, varying subsurface conditions, and indistinct target boundaries. Although visually image-like, radar cross-sectional data fundamentally represent electromagnetic waves, with variations within and between waves critical for identifying anomalies. Addressing these, we propose the Reservoir-enhanced Segment Anything Model, a framework exploiting both visual discernibility and wave-changing properties of radar data. The model initially identifies visually apparent candidate anomaly regions and further refines them by analyzing anomaly-induced changes within and between electromagnetic waves in local radar scans, enabling precise and complete anomaly region extraction and category determination. Real-world experiments demonstrate that the model achieves high detection accuracy ( > 85%) and outperforms existing methods. Notably, it requires only minimal accessible non-target data, avoids intensive training, and supports both fully automatic operation and simple human interaction to enhance reliability. Our research provides a scalable, resource-efficient solution for rapid subsurface anomaly detection across diverse environments, improving urban safety monitoring while reducing manual effort and computational cost.