PATTERN ANALYSIS TECHNIQUES FOR DETECTION AND IDENTIFICATION OF BURIED TARGETS USING SFCW GPR

Abstract

Ground Penetrating Radar (GPR) is considered as an environmental tool for detection of buried objects such asUnexploded Ordnance (UXO) and landmines. Due to their system complexity and high component cost, it was harder to use it. However, over the last decade the cost of RF technologies has decrease considerably, making it more feasible to make SFCW GPR. The basic concept of SFCW based GPR are introduced in the thesis and a critical study and extensive experimentation have been carried out to detect and identify the various targets like -air cavity, a metal sheet, a landmine and a water bottle were buried at different depth and moisture condition. The targets were also buried and tested for detection under plywood of thickness 18mm, 12mm and 6mm. By the application of preprocessing techniques like Hamming Window Filtering, Median Filtering on Bscan image is constructed. To remove clutter from the image, SVD and lCA were applied on B-Scan images. Once presence of targets was detected, classification and identification of targets was carried out. Different techniques were applied for classification. Second, targets are identified using a contextual masking technique which is based on intensity of reflected signal. Lastly, pattern recognition models such as Probability Density Function based classification and Neyman Pearson algorithm along with likelihood ratio test (LRT) method is applied. Use of artificial intelligence methods such as Genetic programming and neural networks is to be applied in future work