Sweeping for mines... with algorithms

Picking a needle out of a haystack is all in a day's work for John Goutsias.

The Hopkins electrical engineer is developing mathematical tools that may ultimately be used to detect above-ground land mines. Such mine detection technology could help protect soldiers entering a war zone, or civilians who want to return to homelands that recently witnessed battle.

For experimental data, Goutsias is using digitized video frames provided by the U.S. Navy. The videos were recorded by cameras mounted on unmanned planes that flew over a test region.

Goutsias must determine whether the signature of a mine or mines lurks within the grainy images. But that is no easy task.

The planes fly relatively high, and the mines are relatively small--perhaps about a foot or two wide, guesses Goutsias who is not privy to that information. "There also might be shrubs, rocks, or trees that look like mines. You cannot see much," he says.

In recent tests, Goutsias's mathematical program was able to detect 94 percent of the land mines in a set of aerial photos.

Copies of digitized video frames on Goutsias's desk resemble smeary Rorschach blots, occasionally speckled by specks of white. Goutsias, who is funded by the Office of Naval Research, has developed a set of mathematical formulas to determine which white flecks are mines and which are most likely something else.

The camera on the unmanned aircraft films each scene through each of six different colored filters, producing six different snapshots of each scene, explains Goutsias. One filter, for example, highlights wavelengths reflected by vegetation. Because of the way they reflect light, land mines tend to show up in more of the images than most other objects.

Goutsias's program flags objects the size of land mines that appear in at least three of the six frames for any particular scene. The program also uses a special technique to reduce "false alarms."

In recent tests, Goutsias's program detected at least 94 percent of the mines in a set of aerial pictures. The program missed about 6 percent of the mines and produced false alarms about 6 percent of the time, reports Goutsias. His method appears to outperform a competing algorithm, which identified only 86 percent of the mines in the same data set, he says. He will present recent results of his studies in Orlando this month during a mine detection workshop at a meeting of the International Society for Optical Engineering. Doctoral student Ulisses Braga-Neto, and former doctoral student Ashish Banerji, now at Hughes Network Systems, helped develop the program.

Over in the Mathematical Sciences Department, assistant professor Carey Priebe is developing a complementary set of mathematical tools to hunt for mines. Priebe begins with the same video data that Goutsias uses. But rather than look for single mines, he searches for mine fields. "Mines don't come in sets of two," remarks Priebe. "They're more likely to be in sets of 10 or 20 or more." Using statistics and probability theory, he looks for particular patterns and clusters in images that could indicate the presence of mine fields.

A 1996 UNICEF report estimated that there are 110 million land mines throughout the world. Most of these older mines were buried or are now covered by rubble and debris; thus, they wouldn't be detected through Goutsias's analysis. His technique is better suited for identifying the above-ground mines that would most likely be used during a future war.

Goutsias is now developing algorithms for spotting underwater mines. --Melissa Hendricks

The silver looking a little tarnished this spring cleaning season? Try electrons.

Jerome Kruger, Hopkins professor of materials science and engineering, is sharing a rare commodity: a scientific home recipe for cleaning silver. The Kruger Silver Method requires only four easy steps using three household staples: baking soda, water, and aluminum foil.

1. Wrap the entire silver object in aluminum foil. Be sure the foil makes tight contact with the silver in at least one place. Make a few cuts in the foil to allow liquid to seep into the space between the silver and foil.

2. Place the foil-wrapped object in a large pot--glass, enameled, or stainless steel--and put it on the stove.

3. Fill the pot with the solution--4 to 5 tablespoons of baking soda per quart of water--covering the object. Let the mixture simmer, gently, for a half hour.

4. Remove the foil. Rinse the silver and dry the heirloom.

The process works, Kruger says, because of an electrochemical reaction. Electrons flow from the foil through the silver object, reducing the tarnish layer on the surface so that it dissolves away in the baking soda solution.

Kruger, an internationally respected expert on corrosion and tarnish, cautions that the method might not create the brightest gleam, so a little rubbing may still be needed. As Kruger says, in true Hints for Heloise fashion: "You can almost wipe it off with a cloth and bring out the shine." --Joanne P. Cavanaugh

Move over, radar

A panel of experts from the Johns Hopkins Applied Physics Laboratory gives the thumbs-up to using the Global Positioning System (GPS) as the nation's sole means of air navigation. Pilots, navigators, and air traffic controllers can safely rely upon the 24 satellites known as the GPS in place of radar and other current land-based navigation means, according to a recent six-month report sponsored by the Federal Aviation Administration and trade and pilot associations.

The FAA plans to phase in GPS for air navigation over 20 years. Advocates hail GPS as an inexpensive system that could absorb the exponential increase in air travel expected in the 21st century, without significant added cost. GPS, they argue, also conceivably could reduce travel times because pilots would not need to fly within range of ground-based navigation devices. Instead, they could travel the shortest path between departure and destination points.

GPS is being adopted by more and more pilots and airports. About 1,000 airports have some degree of GPS capability, according to Elmer Mooring, an APL engineer who led the study group and is also a pilot and flight instructor. "We're just nibbling at the beginning of the new technology," he says.

But critics fear that solar flares, jamming by terrorists or hackers, or television and radio transmissions could interfere with the GPS signals. Another concern is that the U.S. Department of Defense, which provides the GPS, could decide to deny sufficient access to the signals.

"Our study said you're absolutely right to be concerned that there are risks," but each risk could be mitigated, says Tom Thompson, the team's technical lead.

The study group recommended several technical measures to augment the current GPS navigation system. These include adding monitoring systems known as WAAS and LAAS, to detect any errors in the GPS signal and to notify aircraft of any necessary corrections. The team also advised incorporating improved GPS antennae and anti-jamming technology on aircraft.

Most potential sources of interference, such as intentional jamming of the signal, would cause inconvenience rather than disaster, says Mooring. "It's not that jamming would cause a plane to crash," he says. GPS would indicate that the signal is inadequate to allow for landing. Air traffic controllers would then need to divert planes to another airport, just as they do now during inclement weather.

Aircraft will continue to have back-up navigation systems such as radar, Thompson notes.

"I would have no qualms--none whatsoever--" about flying on a plane that uses GPS navigation, says Thompson. "The technologies are basically proven." However, a significant amount of planning and coordination will be required to implement all-GPS navigation, he says.

The GPS study group included representatives from Stanford and Ohio State universities, Raytheon and Boeing corporations, and the sponsoring organizations. --MH

A journal you can count on

First proposed by Hopkins math professor J.J. Sylvester, the American Journal of Mathematics was born in 1878, launching the Johns Hopkins University Press. About 100 subscribers paid $5 for the inaugural volume. Today, AJM is the oldest continuously published math journal in the Western Hemisphere.

So, you're tired of hearing about the Y2K problem. Tell that to your sturdy old computer next New Year's Day. Hopkins information specialists are dispensing online advice to cut through the hype and get straight to easing the headache.

To safeguard your circa 1982 IBM or figure out whether newer equipment is vulnerable at the turn of the century, visit www.jhu.edu/ywk/, a site recently launched by Homewood Academic Computing. The futuristically titled Year 2000 Center gives an overview of the problem and offers tools to analyze hardware and software applications for DOS/Windows and Apple II/Macintosh computers.

On the web page, Hopkins computer experts provide tips--check your BIOS and your peripherals--as well as warnings: Jan. 1, 2000, falls on a Saturday, meaning your first Monday at work in the new millennium could be a particular downer.

The site also offers soothing Words of Relief: "The Year 2000 problem has no single 'fix,' but it is . . . manageable;" and lists answers to Frequently Asked Questions, links to other Y2K web sites, as well as how to reach Hopkins experts accustomed to panic situations. --JPC

RETURN TO APRIL 1999 TABLE OF CONTENTS.