Investigation of the Effects of Buried Object Orientation in Subsurface Target Detection by GPR

Use of ultra-widehand ground penetrating radars (GPRs) have been the most effective approach in subsurface target detection. Proper preprocessing, prescreening, feature extraction, classification and fusion techniques are all needed to satisfy the requirements of high detection probability and low false alarm rates, simultaneously. Dominating ground reflection signals and soil clutter effects including the randomly varying non-planar ground surfaces, inhomogeneous soil nature, pebbles or rocks within the soil, moisture level of the soil and various garbage items in the soil environment are all well-known and broadly-studied problems in the GPR-based target detection and recognition. An additional source of difficulty in this problem is the arbitrary orientations of the buried objects, which may lead to incorrect classification results, and hence increased false alarm rates. In this paper, we will first simulate down-looking GPR signals for cylindrical targets, which are made of perfect electric conductor (PEC) or plastic, buried in soil in different orientations with respect to the air-ground surface. Then, we will remove the dominating ground reflections from the raw data by a suitable preprocessing technique to facilitate the detection of buried objects. Finally, target features will be extracted in the two dimensional temporal-spectral domain by using the Page distribution (PD), which is an energetic time-frequency representation (TFR) technique. The aim of this paper is to investigate the variations in the GPR signals and also in the extracted target features due to changes in the orientations of buried targets.