Hopkins Engineers Use Advanced Math to Find
Deadly Devices in Aerial Pictures
Soldiers are preparing to step onto an enemy beachhead, where the sands may be studded with land mines. Military commanders need to find out where these devices are located. So, before the soldiers arrive, an unmanned plane soars high overhead to shoot video pictures of the beach. There's one hitch, however: from this altitude, you can't tell whether those dark spots are deadly mines or harmless rocks and trees.
To solve this problem, engineering researchers at The Johns Hopkins University have developed a series of complex mathematical steps that allow a computer to filter out unwanted material in the surveillance images and locate the land mines with a high degree of accuracy.
"This algorithm is simple, performs well and requires only an approximate knowledge of the size of the mines," says John Goutsias, a professor of electrical and computer engineering in Hopkins' Whiting School of Engineering.
Goutsias and doctoral students Ashish Banerji and Ulisses Braga- Neto described the new mine detection strategy recently in the IEEE Transactions on Aerospace and Electronic Systems and during conferences organized by the International Society for Optical Engineering. Using minefield video images supplied by the Coastal Systems Station, Naval Surface Warfare Center, the researchers reported a 95 percent success rate in finding devices that had been placed above ground. Goutsias hopes that other researchers will use this mathematical tool as the foundation for new, more sophisticated systems for locating above-ground land mines.
Military leaders and humanitarian agencies alike are anxious to find new ways to locate and disarm these explosives. A 1996 UNICEF report estimated that 110 million land mines were strewn throughout 64 countries, endangering civilians as well as military troops. Many of these mines were activated during conflicts that were resolved years ago. Yet the devices, often costing as little as $3 apiece, still pose a serious threat.
Using an unmanned plane to photograph a potential mine field is one way to avoid risking a pilot's life, Goutsias says, "but unless you can come in very close to the ground, it's difficult to see whether there are mines. First, the mines are relatively small. And second, they're usually in areas that have a lot of vegetation, where it's difficult to discriminate among shrubs, trees and mines." To overcome this hurdle, military teams often install a six- segment filter that spins in front of the camera lens aboard the plane. Each segment is a different color, allowing different optical frequencies to pass through. Because there are six filters, the camera produces six different images of the same scene. Each filter causes certain parts of the picture to stand out. For example, one filter may cause vegetation to show up more prominently because of the way chlorophyl reflects light.
The algorithm developed by Goutsias, Banerji and Braga-Neto enables a computer to use size and shape restrictions to disregard trees and other large objects that are unlikely to be a mine. But what about a rock or a shrub that is roughly the size of a mine? To eliminate these, the process compares the six filtered images of the same scene. Rocks and vegetation reflect light differently than metal or plastic mines, and they show up less often in the images. "If an object of the proper size appears in at least three images, the system decides it is probably a mine," Goutsias says.
Following the steps developed by the Hopkins researchers, a computer can find the above-ground mines in an aerial picture in less than a minute, Goutsias says. The process needs further refinement, but the Hopkins professor believes it will produce results that are superior to existing aerial mine detection systems.
Goutsias' research in mine detection has been funded by the Mathematical, Computer and Information Sciences Division of the Office of Naval Research, in collaboration with the Coastal Systems Station, Naval Surface Warfare Center in Panama City, Fla..
Goutsias completed his undergraduate studies at the National Technical University of Athens in Greece. He earned his master's degree and doctorate in electrical engineering at the University of Southern California in Los Angeles. He has been a faculty member at Johns Hopkins since 1986. The professor has just launched a new line of research aimed at detecting mines hidden underwater.
Color slide of Goutsias and black and white illustration of mine detection system available; Contact Phil Sneiderman
Go to Headlines@HopkinsHome Page