Buried Objects Segmentation and Detection in GPR B Scan Images

Abstract

Identificationof buried objects through Ground Penetrating Radar B scan (GPR-B) images needshigh computational techniques and long processing time due to curve fitting orpattern recognition methods. In this study, an efficient and fast recognitionsystem is proposed for detection of buried objects region. Previously, GPR-Bscan images of objects with different shapes in various depths were obtained byusing gprMax simulation program. The detection process is categorized into foursteps. The GPR-B images are transformed at first step. Then, they arethresholded to obtain potential buried object regions. Third step of the systemis hough transform in order to eliminate ground surface. Finally, an estimatedregion analysis is performed. The results show high performance with fullyautomatic segmentation. The processing time for detection of buried object isin the range of 1.234 - 2.232 seconds. It can be observed that this technique isfaster than other studies in the literature. Consequently, it may be used inreal time applications for GPR devices.