Buried target detection with ground penetrating radar using deep learning method

Deep learning has started to outperform its rivals over the last five years, due to its capability to automatically find the features in the data, and classify them. In this study, deep learning is used to detect a buried target collected by a ground penetrating radar (GPR). The GPR data is generated by the GprMax simulation program, and a deep learning model of two convolution and two pooling layers is proposed to classify this data. The proposed model is trained with two classes, with a hundred targeted targets and a hundred non-targets. At the end of the training, the resulting features were examined in each layer of the deep architecture. The initial results presented in this study emphasize the advantages of deep learning over traditional classification methods, since it allows for high classification rates without the need for feature extraction.