Performance assessment of SVM-based classification techniques for the detection of artificial debondings within pavement structures from stepped-frequency A-scan radar data

GPR Monitoring of Artificial Debonded Pavement Structures throughout Its Life Cycle during Accelerated Pavement Testing

“…ventional amplitude ratio test (ART) as the ratio between the reflected wave (RW) and the direct wave (DW), namely, ART = RW/DW [15,18]. Then, the data processing basically consists of automatically picking the maximum (or minimum) amplitude of the two latter echoes with some existing commercial radar software.…”

“…Then, the analysis of the amplitude of the surface direct wave itself was carried out, as characterizing the first layer is also a subject of interest. The next step in detection debonding requires to compute, for each A-scan, the conventional amplitude ratio test (ART) as the ratio between the reflected wave (RW) and the direct wave (DW), namely, ART = RW/DW [15,18]. Then, the data processing basically consists of automatically picking the maximum (or minimum) amplitude of the two latter echoes with some existing commercial radar software.…”

“…Unfortunately, no visible surface degradation was observed at the end of this first stage of experiments. As a result, the ongoing experiment has prompted 6 years of several different actions, including internal and international research projects [13][14][15]. Recently, a final series of loading has occurred, with the support of three national contributions, to complete this experiment, allowing to reach an advanced level of degradation requiring repair from the point of view of the engineers managing road networks.…”

A Review of GPR Application on Transport Infrastructures: Troubleshooting and Best Practices

“…Methodologies based on linear prediction and supported by vector regression have been proposed and tested using a 10 GHz bandwidth antenna in an accelerated pavement testing facility [236], providing promising results in the detection of thin debonding. Another proposal was the application of a support vector machine to the A-scans, allowing for the detection of horizontally stratified debonding and geotextiles between the two upper pavement layers [237].…”

“…However, Zhang et al [45] propose the use of more properties of the dataset to detect wet zones, considering statistical properties that are affected by moisture, selected with linear discriminant analysis. The comparison of results from GPR tests and passive microwave radiometry demonstrates that GPR is able to detect water content at higher depths, enabling the detection of water table in the soil under the structure [237]. Nevertheless, the identification in the B-scans of the anomaly associated with water table can require different seasonal measurements in order to determine the permanent and variable reflectors, which allows one to distinguish between natural layers or pavement structures, and water table.…”