Classification of Targets Using Time-Frequency Analysis of GPR Data

Abstract

Ground penetrating radar (GPR) is a potential method in demining techniques due to its capacity to identify plastic and metal-cased antipersonnel landmines. However, detecting landmines with GPR is difficult because other subsurface reflectors, such as stones, or metallic debris, can interfere with the detection process. This paper analyzes and evaluates a target discrimination approach on both synthetic and measurement data. It is based on significant features collected from the time-frequency domain of 1-D GPR signals using the continuous wavelet transform (CWT) and the Wigner Ville distribution (WVD). These extracted features are then fed into the Neural Network (NN) classifier for effective classification. The proposed algorithm is evaluated using radar data gathered in controlled laboratory settings utilizing stepped frequency continuous wave (SFCW) GPR. The outcomes and performance metrics of this experimentation showcase the efficacy and potential of the developed approach in landmine detection and discrimination