Magnetometer-Based Magnetic Anomaly Detection, 14-R8064Printer Friendly Version
Inclusive Dates: 04/27/09 08/27/09
Background - In a 2008 Broad Agency Announcement, the U.S. military solicited proposals for new methods of detecting Improvised Explosive Devices (IED) buried in and along roadways. SwRI submitted a proposal in response to that solicitation describing a detection approach based on using a vehicle-borne array of sensitive magnetometers to map and display the distribution of overturned soil in the roadway. SwRI was asked to participate in a blind test of buried IED detection at Yuma Proving Ground in the summer. To get a field-capable system ready and tested in time, SwRI used internal research funding to survey native, graded, and trenched soil in an undeveloped area of SwRI's grounds.
Approach - Although SwRI's proposal was based on a development effort for its Rubidium Vapor Magnetometer, to field an array quickly SwRI elected to use commercial-off-the-shelf (COTS) magnetometers of roughly equal sensitivity to develop the measurement technique and signal processing for trench detection. SwRI researchers used vertical and horizontal gradients of the total field, and the SwRI detection algorithm was based on a maximum likelihood comparison of the gradient signals to the expected gradients from a buried, magnetically permeable region of Earth.
Accomplishments - The magnetic field anomalies caused by the removal and replacement of soil are extremely small, about 1 to 5 nanoTesla (about 1/10,000 of the Earth's ambient magnetic field). Thus, several signal processing steps were used to bring the anomalies to light. The magnetometers were arrayed on a nonmagnetic cart, two laterally near the earth and a third one meter above the others, as shown in Figure 1. The cart was then taken out to survey the selected area, which was chosen because it was undeveloped and had an existing dirt track that was useful for grading the primitive roadway. Baseline magnetic field data were taken at approximate one-foot intervals over a 100-foot by 50-foot area that crossed the dirt track, which became the roadway. The track was graded to simulate a two-lane dirt road, and a second set of data was taken over the same area as the baseline. Next, three trenches were dug part-way into the graded road to simulate what would be done to bury an IED trigger cable and device. These trenches were 1 foot, 2 feet, and 3 feet deep by 2 feet wide, and ran from 12 feet off the roadside 8 feet into the road. A second set of data was taken over the area. The track was then graded to simulate a two-lane dirt road, and a third set of data was taken over the same area as the baseline. To discriminate SwRI's signal from the clutter, a model of what the signal from a trench should look like was used to run a novel maximum likelihood estimator (MLE) on the data. These processed maps, illustrated in Figure 2, show the trenches more clearly than the gradient data alone. The relatively large amount of clutter from the uncleared area was a surprise. Several anomalies of more than 5 nT size were found in the survey area; even using the MLE, these were sufficiently large to set off a simple detector based on that output.