Technologies that provide new solutions to the diverse detection and discrimination problems of unexploded ordnance (UXO)-contaminated land sites are needed.

Increased detection probabilities, reduced false alarm rates, and better characterization and discrimination of subsurface UXO can significantly reduce the residual risk and costs of response and allow more rapid and safer transfer of lands. Over the past five years, substantial investments have been made toward the goal of developing systems that can efficiently screen clutter from ordnance. However, current systems cannot meet the goal of nearly 100 percent detection with low false alarm rates. Results have estimated that one of the most important factors in using UXO sensor data to characterize buried targets and discriminate between UXO and clutter is the precise knowledge of sensor location and attitude while the data are being collected. This project aimed to provide the means of obtaining that information with handheld UXO sensors. Handheld sensors are required for areas inaccessible by vehicle or man-portable systems because of terrain or vegetation.

The objective of this project was to develop an inexpensive, robust way to accurately determine the trajectory of a handheld UXO sensor as it is swept about above a suspected buried UXO item.

In principle, sensor position and orientation can be tracked using an inertial measurement unit (IMU) that measures accelerations and angular rates along three orthogonal axes. Whether a compact, inexpensive, rugged IMU based on micro-electromechanical systems (MEMS) technology is sufficient was a main focus of this project. Another point of focus was the development of processing procedures that combine the trajectory with the sensor output stream to characterize the target for discrimination between UXO and clutter.

This project integrated a compact, rugged inertial measurement unit (IMU) with a handheld UXO sensor. The integration involved the use of a Hidden Markov Model (HMM) to estimate the motion state of the sensor (e.g., turn, middle of sweep over target). Adaptive IMU noise suppression was controlled by the HMM. The HMM also controlled procedures for using the UXO sensor output as an external aid to the IMU to compensate for bias drift. A post processor combined the position information from the corrected IMU with the sensor output stream to characterize the target and discriminate between UXO and clutter.

The researchers demonstrated that a compact (7 cm square), lightweight (less than 400 gm) Honeywell HG1900 IMU can provide sufficiently accurate positioning to support accurate estimation of target parameters with the EM61- hand-held (HH) sensor. The Honeywell unit costs approximately $10K. The researchers also tested a less expensive (approximately $4K) unit, the Crossbow IMU400CC-100, but found that it did not provide sufficiently accurate positioning. The major problem with the Crossbow unit appeared to be that its angle rate noise and stability are not sufficient to accurately track attitude and keep gravitational acceleration from contaminating the horizontal acceleration channels. The angular rate random walk spec for the Crossbow unit is 2.25 deg/(hr^1/2), while the corresponding spec for the Honeywell unit is 0.1 deg/(hr^1/2).

The researchers found that the response characteristics of the EM61-HH can significantly affect the accuracy of target parameter estimates when the sensor is moving about over the target. This occurs because the sensor output is integrated with an analog filter that both shifts and distorts the sensor's response. The researchers measured the filter response characteristics and included it in a dynamic forward model for the EM61-HH. In order to achieve discrimination-quality accuracy, the dynamic model has to be used when inverting data collected as the sensor is swept about over a target.

In a few of the test cases with the Honeywell, the researchers had an uncorrected drift of several centimeters relative to ground truth provided by a laser (ArcSecond) positioning system. This type of drift can be corrected using an adaptive processing approach which includes the EMI sensor data in the track estimation process. The trajectory errors arise because of errors in estimating the IMU biases. By including the bias parameters in the EM inversion procedure, the best track estimate along with the target parameters (location and principal axis polarizabilities) can simultaneously be determined.

Output from a small IMU sensor is co-processed with handheld EMI sensor output to produce an accurate EMI anomaly map.

The improved positional accuracy information of a HH UXO sensor will enhance its use in reliably discriminating UXO from clutter. Cost savings can be achieved by eliminating the excavation of non-ordnance items.