Automatic Detection of Buried Objects Depending on Ground Penetrating Radar Images Using Convolutional Neural Networks

Abstract

Ground Penetrating Radar (GPR) is one of the technologies that has proven its efficiency in detecting buried objects by using electro-magnetic waves.

However, this technology suffers from the problem that accurately identifying buried objects through GPR images remains a challenging task that heavily relies on the experience of the GPR device operator. So, the results of interpreting images may be unreliable, and require hard work and a long time for interpretation. The aim of this research is to help GPR device operator overcome this problem by building a Convolutional Neural Network (CNN) model that automatically detects the hyperbola signature reflected from the buried target (buried pipe in soil as case study) in GPR images from type of B-Scan (Brightness -Scan). There is a challenge in providing the necessary dataset for deep learning models in the field of GPR technology, due to the scarcity of real GPR data discussed by experts in this field. Also, obtaining labeled GPR images is not always available. Therefore, in this research, two datasets consisted of simulated GPR B-Scan images using the gprMax simulation software, and real GPR B-Scan images were created, which were used to train and test the CNN model. The proposed model achieved accuracy = 100% on the validation dataset and accuracy = 97.5% on the test dataset, which represents invisible data to the model. This demonstrates the high performance of the proposed model, and thus the proposed CNN model can be used as a supporting system for detecting buried objects by GPR device operators who do not have sufficient experience in the field of GPR technology.