Fundamental Shape Discrimination of Underground Metal Object Through One-Axis Ground Penetrating Radar (GPR) Scan

Abstract

Ground Penetrating Radar (GPR) was used in this research to detect or recognize the buried objects underground. Hyperbolic signals formed by datagram of GPR after detection the buried objects which quite similar to each other in term of metal shapes. The research was tested on the metal cube and metal cylinder by using the A-scan of GPR. There are steps in this signal processing step which are pre-processing step, feature extraction, and classification process. The segmentation process hyperbolic signals were segmented one by one and normalize from the negative to positive signals. The hyperbole from the metal cylinder and metal cube that had been buried in the ground is differentiated using four features of their respective A-scans which are found the maximum value of amplitude signal graph, the number of peaks in the signals graph, skewness, and standard deviation values. Finally, the classification process used learning algorithm of Multi-Layer Perceptron (MLP) was a test on Bayesian Regulation Backpropagation (BR) was given the highest accuracy, 98.70% as a classifier to classify the metal shapes which are a metal cube and metal cylinder.