Hazardous Zone Identification by Spectral Analysis of Ground Penetrating Radar Response

Ground penetrating radar (GPR) has evolved over the years as a profound sensor-based investigation technique operating in a wide range of frequencies ranging from 250 to 1000 MHz and utilizing interaction of electromagnetic waves with subsurface to obtain a pseudo-image of the strata. Often due to constructional negligence and poor standards of construction, the pavement so constructed lacks structural strength and have poor compaction of construction material resulting into air pockets. During rainy season, water seeps into the pavement resulting into settling of sand and gravel underneath the pavement. Hence, when a heavy vehicle crosses over the road patch, it results in subsidence of that patch leading to casualties and property damage. In the present study, a system is developed for classification and identification of hazardous zones on a GPR image that can result in subsidence during rainy season. The system works on extraction of features from GPR response using discrete Fourier Transform. The GPR data collected at two sites in IIT Roorkee campus using 1 GHz antenna are fed to the support vector machine for training the classification system. It has been observed that presence of clay and trapping of water beneath the pavement resulted in subsidence of the patch. From the study, it can be deduced that the GPR can be used as a non-destructive tool for hazardous zone identification and pavement fault detection. Index Terms--Discrete Fourier Transform (DFT), ground penetrating radar (GPR), hazardous zones, Non-destructive evaluation, support vector machine (SVM)