Physics-based Data Fusion for Uxo Characterization

ENSCO, Inc. is investigating a novel approach to data fusion of magnetic and ground-penetrating radar (GPR) data to characterize previously detected buried targets. The method exploits a single model of the subsurface to predict both magnetic and GPR data observations. By determining the best single model that constrains both data sets, the optimum identification is computed. Using an a priori database of specific potential targets, the quality of fit between measured and predicted data for each potential target is calculated. The potential target for which the model best fits the data indicates the identification of the target. Vertical component magnetic gradient data is modeled using a finite-length dipole model (as compared to the more common point-dipole model). GPR data is evaluated both in terms of diffraction geometry and radar cross-section. Quality of fit between model and measured data is evaluated via a non-dimensional metric. This paper reports on a work in progress that will be completed during 1998. INTRODUCTION Characterizing the identity of a buried target has proven to be a very difficult task. Conventional unexploded ordnance (UXO) detection surveys typically produce numerous anomalies that, upon excavation, prove not to be of interest. Therefore, the cost of UXO remediation would be greatly reduced if we could reliably determine the character of buried targets. The hazards of UXO excavation would also be reduced if explosive ordnance disposal (EOD) technicians had reliable knowledge of the type, condition, and orientation of the UXO item in the ground. The lack of success to-date in developing a capability to characterize UXO is due, in part, to the reliance on magnetic sensors for the majority of UXO detection work. Using these same data to approach characterization has been only marginally successful. It seems clear that other, or additional, sensors will be need to characterize UXO. However, a key difficulty in applying multiple sensors is combining disparate data to produce a single answer of the target identification. This paper presents an approach ENSCO is investigating to combine magnetic and ground-penetrating radar (GPR) to identify UXO. The data fusion framework is suitable for incorporation of other data types, such as induction electromagnetics. This project is being conducted as part of the Jefferson Proving Ground (JPG) Phase IV Advanced Technology Demonstration (ATD).