DURHAM, N.C. -- New techniques for analyzing signals from the kinds of land mine detectors now in use could for the first time enable mine removal teams to discriminate real mines from other buried clutter, according to Duke University electrical engineers developing the techniques.
The mathematical methods for analyzing signals from electromagnetic land mine detectors could be programmed into microchips to create "smarter" systems, the researchers said. These innovative signal-processing techniques could be programmed into microchips to tap signal information disregarded by designers of currently used detectors.
"The current detectors are the only ones soldiers now have, so if we can improve them we can really make an impact," said Lawrence Carin, a Duke associate professor of electrical and computer engineering, who is leading a $6 million U.S. Army funded multi-university research initiative (MURI) to spearhead new approaches to electronic land mine detection.
More effective detection technology would reduce the high cost of finding and neutralizing real land mines among a much larger array of buried false "targets." That per mine neutralization cost is currently estimated to be as high as $400 for devices Iraqis implanted in Kuwait during the Persian Gulf War.
"It would take about 1,100 years, at current clearance rates, to remove all the land mines in the world," said research colleague Leslie Collins, an assistant professor of electrical and computer engineering at Duke. "That's assuming no more are laid," she added. "But they are laying more mines than you can remove in a year."
Several Duke land mine detection investigators described their ongoing work in this area at a Department of Defense "clutter conference" held Oct. 29-30 at the Institute for Defense Analyses Science and Technology Division in Alexandria, Va. Other presentations are scheduled next year at meetings in Orlando, Fla., Greece, France and Israel.
The project also involves researchers at California Institute of Technology, Georgia Institute of Technology, Ohio State University and Stanford University as well as Oak Ridge National Laboratory and three industrial partners, EG&G of Albuquerque, N.M.; Hughes Aircraft Co. of Malibu, Calif.; and Northrop Grumman Corp. of Baltimore, Md.
In an interview, Carin said the "electromagnetic induction" (EMI) devices now widely employed to locate mines are essentially the same as the metal detectors used to find lost coins in beach sand or buried artifacts on Civil War battlefields.
EMI technology "has been around since the 19th century," he added. In fact, doctors unsuccessfully tried to use an early version to pinpoint the assassin's bullet that felled President James Garfield. That attempt failed only because of interference by metal springs in Garfield's bed, Carin wrote in one of several related research papers recently submitted to Institute of Electrical and Electronic Engineers (IEEE) publications.
In EMI, electrified metal coils in the detector's tip create a magnetic field that can penetrate the ground to reach a buried object. Objects that respond to magnetic fields -- especially metals -- then perturb the induced field. That field disruption, which varies according to the nature of the target object, is then sensed by the detector as a returning signal.
Operators can adjust an EMI detector's sensitivity so it sends them a tone through a headset. When properly adjusted, the detector produces an audible tone only when the return signal reaches an optimal strength that indicates the presence of an object of interest.
Researchers around the world -- including those at Duke -- are now investigating futuristic new land mine detection approaches, but EMI-based devices "are now the only working sensors in the field today," Carin said.
The major problem with current EMI detectors is that they cannot discriminate metal mines from other metal objects that may emit signals of similar intensity. So, the devices may wrongly identify an array of other debris as possible mines. Such sham signals mean that mine deactivation and removal teams must endure many false alarms as they go through the harrowing and dangerous ordeal of probing and digging the ground around a potential target.
"Nobody has really ever cared about discrimination before," Carin added. "It's only relatively recently that people even knew why EMI worked." So Carin decided to go back to the basics in his research. Exploring the "underlying physical principles" behind electromagnetic induction, he investigated in detail how a target's shape, size and composition affect the resulting signal.
He realized, for example, that buried objects with a certain "rotational symmetry" are more likely to be land mines. More significantly, Carin discovered that the magnetic energy from objects being stimulated by the sensor contains a rich array of previously untapped information.
In one set of studies, the engineers analyzed how the signal decayed with time, like the decay of the sound of a pinged bell. "In this case, we ping the object with a short pulse of electromagnetic induction," he said. "Then we 'listen,' in an electromagnetic sense, to how the energy decays as a function of time."
Carin is also exploring how other useful information can be extracted by pinging mines with different frequencies. He uses a "wideband" EMI land mine detector under development by Geophex Ltd, a Raleigh, N.C., firm. This wideband technique transmits the initial pulse in several different frequencies instead of just one.
Collins is working closely with Carin to develop mathematical rules, called algorithms, that can analyze the complex incoming data from both "time and frequency domains" with the help of "Bayesian" statistical methods.
The Bayesian approach, which can be incorporated into a data processor, inputs the available information from detectors and then calculates the probability that a buried object is, say, a land mine as opposed to a collection of nails.
Carin and Collins have tested their algorithms using EMI sensor data recorded at an Army land mine sensor testing site, where mines and other objects are buried at precisely known locations.
"We've been able to show, whichever route you go, you can do a better job of rejecting things like nails and identifying buried mines," Collins said. In fact, they've found their approach produces "a "five-fold decrease in the false alarm rate," she added. "Normally, for every mine you average finding 100 pieces of clutter. So, effectively, what this decrease means is instead of finding 100 pieces of clutter you'd only find 20 or 25."
Their results are "really starting to make some ripples within the U.S. Army, because it's so much better than what is currently being done in the field," Carin said.
Collins said an advanced detector incorporating their algorithms on a digital processing chip would no longer beep every time a detector senses the presence of buried metal; only when the sensed object is a likely land mine.
Carin is also investigating the use of EMI detectors to sense the presence of plastic mines, which are unresponsive to magnetic fields.
In such experiments, researchers would throughly soak the ground with water to boost the conductive properties of the underlying soil. Then the Duke engineers could use their analytical techniques to look for simulated plastic mines in "conduction voids," where an EMI signal is absent. "A plastic mine is a very poor conductor," Carin said.
Meanwhile, Erol Gelenbe, a professor who chairs Duke's electrical and computer engineering department, has invented another algorithm to help EMI sensor operators weed out false from real targets.
Called the "delta technique," Gelenbe's mathematical method searches for characteristic cone-shaped patterns in graphical representations of the reflected energy of EMI signals. His analyses show these cones tend to be present near mines. The technique can mean 150 percent fewer false alarms, he said. It could be programmed into a portable computer for use in the field, he added.
"It's a way of cleaning out clutter and identifying those points which are significant candidates for being the locations of mines," Gelenbe said. "But this technique, like any other, does not guarantee that what remain are actually mines; it simply reduces drastically the cost of de-mining or of finding unexploded ordinance."
Other MURI studies are exploring the use of synthetic aperture radar, ground penetrating radar and ground vibrations, and infrared sensing. Another would use a combination of ultrasound, microelectromechanical (MEMS) analyzers, and an odor-sensing microchip to detect explosives within buried mines.
###