Processing Ground Penetrating Radar Data To Improve Resolution Of Near-Surface Targets

Although unprocessed ground penetrating radar (GPR) data are capable of producing detailed images of the shallow subsurface, these images may be difficult to interpret reliably. Here, we<br>investigate characteristics of GPR signals and various techniques to enhance GPR images. The processing steps are tested on GPR datasets acquired at a well-controlled site where two concrete blocks, simulating archaeological structures, are buried less than two meters from the surface. Short range GPR signals often possess a low-frequency component (commonly referred to as a ttwowtt) that causes amplitude distortion along an individual trace. Characteristics of this noise are considered in light of its physical cause, i.e., saturation of the receiver electronics. Three filter techniques were tested to remove the wow: (i) residual mean filtering, (ii) bandpass filtering, and (iii) residual median filtering. We found that the residual median filter performed best. Additional signal processing steps included static corrections, normal moveout (NMO) corrections, migration and eigen filtering. Static corrections compensate for drift of the time-zero sample that occurs during acquisition. NM0 corrections remove spatial distortion due to source-receiver offset. Migration focuses diffracted energy and corrects dipping reflectors thereby improving spatial resolution. Eigen filtering removes the coherent direct arrivals to improve the detection of near-surface features. Application of these processing techniques significantly improves the quality of the GPR images from the test site, both in terms of the positioning of events and resolving capability. By testing various enhancement techniques on data collected at wellcontrolled sites, a better understanding is gained of their relative benefits and the ultimate capabilities of GPR as a shallow exploration tool.