Hyperbola Recognition from Ground Penetrating Radar Using Deep Convolutional Neural Networks

Abstract

Ground penetrating radar (GPR) has been recognized as a useful non-destructive testing and imaging tool for subsurface exploration. GPR data can be recorded at a high-speed in a continuous way. The automatic detection and interpretation of huge amount of GPR data is desired, but challenging. In this paper, we propose to use a model based on deep convolutional neural networks for automatic detection of hyperbola in GPR images. Compared to other machine learning models applied to GPR images, the proposed method needs less preprocessing steps, and does not require edge detection, segmentation and support vector machine (SVM) classifiers. The proposed hyperbola detector is used to detect the buried objects and laboratory experimental results validated its performance. The scores of the proposed hyperbola detector for hyperbola recognition from GPR images are marked.