Trend deviation analysis for automated detection of defects in GPR data for road condition surveys

This paper presents a novel approach for automated detection of defects and structural changes in GPR data acquired in HMA (Hot Mix Asphalt) road surveys. Unlike the majority of existing approaches for road GPR data processing that are mainly used for extraction of layer profile information, the proposed method focuses on automated identification of significant deviations in subsurface structure and material properties. It is based on the detection of variations in intensity trends of longitudinal lines of interpolated B-scans that are characterized by deviations above a defined threshold. The outputs include mapped defects and deterioration areas together with the locations of detected changes in road structure design.