Buried object detection and analysis of GPR images: Using neural network and curve fitting

Abstract

Recent live campaign applications involve the realtime location and identification of buried Improvised Explosive Devices (IEDs) and buried fusing mechanisms for the needs of national security. Ground Penetrating Radar (GPR) is an instrument used in the construction of under ground images. In principle, images of subsurface objects such as mines and pipes may be detected and potentially measured. Noise and clutter are the influential irregularities that are present during GPR raw-data collection where the sampling rate is 8+ frames per sec. Preprocessing techniques on this voluminous data has been proposed. The reflection from mines or pipes in the ground is characterized by a hyperbola on the under ground radar image. The work is focused to simplify the interpretation of the hyperbolic pattern found in GPR image and estimate the position of the objects using neural networks and curve fitting techniques. We devise an efficient dynamic runtime buried object detection algorithm and verify results.