Research in Progress

Problems of commercial aircraft security, contraband detection, and humanitarian demining appear on the surface to represent the application of widely disparate technologies. There is, however, an underlying set of methodologies which makes it possible to apply similar scientific techniques to all three.

My research in the last six or seven years has included work with the FAA on developing new technologies for the discovery of hidden explosive devices. Terrorism and the problem of discovering clandestine explosive devices have led to many developments in explosive detection that may have applicability to landmine detection. The aircraft security problem exhibits at once the worst problems of both military and humanitarian demining. It requires very high detection probability (as does the humanitarian demining problem) as well as the high speed of military counter-mine warfare.

The problem of landmines and their effect on civilian populations well after wars are over has been of great concern to humanitarian organizations for years. Unfortunately, despite the horrendous litany of their effects on the civilian populations and economies of some of the most impoverished countries of the world, until the past few years most of the victims were out-of-sight and out-of-mind of the developed world.

Now, with the December 1997 signing in Ottawa of a convention essentially banning anti-personnel mines, the issue of mines and their removal has become considerably more visible. The treaty not only forbids the use of these mines, but also obliges the signers to ensure the destruction of mines in areas under their jurisdiction within 10 years. Other signers in a position to do so are also to provide assistance.

It should be noted, that the United States did not sign the agreement, arguing that current U.S. policy is to employ only self-destructive/self-deactivating mines. The observed failure of these is about 1 in 10,000 (about this fraction will not deactivate after the required time of several hours to several days) but will become inert after about 60 days. Although it may be argued that this does not create a future humanitarian problem, the non-discriminate nature of mines is still an issue. Further, the "moral firewall" argument still remains since it is clear that some states or organizations that lack U.S. technical capability to produce such mines will produce mines up to their own technical capacity.

With the enhanced visibility of the demining problem has come new interest in methods for landmine detection and in particular to methods which may be appropriate for clearing the huge number of mines already planted. It is this that has sparked a resurgence in technology development for demining, including by several groups at MIT.

The usual dictum in real estate "location, location, location" forms the basis for most of the technology developed for mine detection. The deminer seeks to locate the mines in a given area by various relatively simple technologies and then later removes or destroys the mine in place. (U.S. military doctrine requires destruction of the mines in place rather than removal.) Still, the problem of humanitarian demining is only superficially similar to the military counter-mine problem.

The military problem is to detect and destroy enough mines so that the dangers from mines are comparable to the other hazards of military operations and, since time is important, to do so in as rapid a manner as possible. For humanitarian demining, we have the contradictory requirements of very high detection probability (the UN has set 99.6% clearance as a goal and now "100% removal" has been discussed) and relatively low cost (although what constitutes "low cost" is not very well defined as it is also necessary to include opportunity costs associated with the removal of farm land from productivity and the inability to resettle refugees). Fortunately, the deminers have the luxury of choosing both the time of day and duration for operations and that may make possible the success of the first two requirements.

The technologies currently employed for explosive detection range from magnetic detectors (useful for weapons), planar x-ray scanners with sophisticated use of multiple energy x-ray beams, computerized tomography (CAT) scanners, various nuclear techniques, and highly sensitive chemical "sniffers," both canine and non-canine. The most powerful scanners currently being deployed use high-speed x-ray tomography (basically high-speed medical CAT scanners) to produce three-dimensional density maps of the contents of a bag.

Yet despite the sophistication of the $1M+ CAT scanners, the false alarm rate of these technologies can be as high as 30 percent. The reason for this lies not only in the complexity of the contents of suitcases (a trip to the FAA Technical Center where the contents of lost bags are examined and classified is an enlightening experience; almost anything that fits into a bag will eventually be found in one) but in the fact that the x-ray methods don't detect explosives directly, but only measure apparent density and shape. Nuclear techniques which directly measure elemental composition and thus directly characterize a suspicious object, are still seen as a second tier detector rather than as a primary screener, and only then if their cost can be reduced. Still, the ability to confirm the presence of an explosive through its composition is a powerful tool.

Is any of this applicable to the removal of mines from Angola or Mozambique or Cambodia? Probably not directly, but there are important lessons which we can learn from the aircraft security problem: First, the importance of false alarms, or rather the lack of false alarms, and second, the importance of characterization of the anomaly. High false alarm rates simply convince the user that their equipment is basically useless and that the cure is to turn down the sensitivity so as not to be burdened by false alarms. Of course for the average airport baggage inspector, with odds of more than a million-to-one against there being a bomb in a given piece of baggage, this response is not surprising. In the case of mine detectors, field false alarm rates of several hundred times the actual detection rate have been reported.

Detection methods which only locate anomalies and which cannot determine whether they are mines are saddled with very high false alarm rates. Magnetic detectors have had very high ratios of alarms to actual mines and may in fact be useless in an environment where much artillery fire has occurred. A single 105-mm artillery shell can produce as many as 3000 pieces of metal that can be detected by sensitive metal detectors. Even the addition of shape discrimination, i.e., attempting a crude form of imaging, has not been demonstrated to have any particular advantage in the field, although new developments may improve performance.

Another approach has been the use of scattered x-rays. X-rays are preferentially scattered by low atomic number materials characteristic of explosives and of non-metallic mines, and thus this method is one that has been proposed as a detection scheme for plastic mines. Early lab tests have shown promise, but the realities of field tests with complicated and cluttered backgrounds still must be demonstrated.

Given all of this, is there some clever piece of technology that can cure the scourge of landmines? Probably not. The lack of success of the "War On Cancer" showed how difficult it is to cure diseases despite vast efforts, but there are undoubtedly many methods yet unexamined. Perhaps some will attack one or more parts of this problem and someday reduce the death toll, returning to people the ability to use their land in confidence without fear of injury. Hopefully some of the collective knowledge and experience of the MIT faculty will find them.

For more information, Web sources are the International Red Cross, <http://www.icrc.org>, in the Issues and Topics section, the U.S. Army humanitarian demining site, <http://www.demining.brtrc.com/> and the Demining Technology Center, <http://diwww.epfl.ch/lami/detec/>.