Abstract:
This dissertation deals with application of Stepped Frequency Continuous Wave Ground Penetrating Radar for real time detection and classification of subsurface targets. Rohde and Schwarz provided FSH4 was used in Vector Network Analyzer mode and was interfaced with a computer to enable real time target detection and classification. A double ridged ultra wideband horn antenna (R&S HF 907) was used to transmit as well as receive EM waves.
Three targets- an air cavity, a metal sheet and a water bottle were buried at various depths under different volumetric moisture levels of soil. By the application of preprocessing techniques like Hamming Window Filtering, Median Filtering on A-Scan GPR data range profile was generated. ICA clutter removal technique was applied on B-Scan images generated by clubbing 30 A-Scans to look for probable presence of targets.
Thereafter, application of postprocessing techniques like back projection and Hough Transform was used to enhance target visibility andconfirm the presence of targets at various depths and under various moisture levels of soil. Velocity correction was also applied to locate targets at exact depths they were buried at.
Once presence of targets was detected, energy density spectrum was generated for each location. This energy density spectrum was then used to classify the targets by the use of neural networks for pattern classification. Contextual masking for successive identification of targets was also used to bolster the classification ability of the system.
