Anomaly Detection Method With Rebar Response Suppression for Microwave Ground Penetrating Radar Pavement Inspection

Abstract

Unsupervised anomaly detection analysis using rebar response rejection is presented for microwave ground penetrating radar (GPR)-based pavement inspections. Various approaches have been used to detect cracks, water, or corrosion in buried objects in the GPR model using synthetic aperture radar (SAR) image. However, this does not always accurately identify an anomaly state because the SAR image largely depends on the selected propagation model (e.g., relative permittivity of background concrete media) or suffers from unnecessary responses such as those from rebars. To address this issue, this article first introduces an effective clutter rejection scheme focusing on the rebar response, using transfer-function-based signal extraction to identify anomalous responses from the boundary between the asphalt and its floorboard. Moreover, we introduce several unsupervised anomaly detection algorithms for time–frequency response data if a large part of the investigation area has normal reflection responses. We performed experimental tests on data from real roads in need of repair to validate that our approach can detect internal anomalies in asphalt or its floorboard.

Keywords

• Anomaly detection
• clutter rejection signal processing
• ground penetrating radar (GPR)
• microwave pavement inspections
• time–frequency (TF) analysis
• unsupervised machine learning