Several novel methods of detecting landmines and other buried explosive devices have been proposed and are subjects of continuing research. The use of trained dogs has been shown to be an effective method of detecting landmines, but the use of dogs is costly and can present enormous logistical problems. The objectives of the present research include, variously, supplanting the use of dogs; providing probabilities of detection sufficiently high for humanitarian purposes; enabling safe, rapid detection at low false-alarm rates as needed for military operations; and/or minimizing costs sufficiently to enable widespread application.

One novel method that is suitable for humanitarian detection and is relatively inexpensive involves the use of bacteria that would emit light, indicative of the presence of mines, that could be observed visually or could be measured. More specifically, the bacteria would be genetically engineered to luminesce or fluoresce in the presence of explosive vapors or other characteristic compounds that leak or evaporate from mines and are thus present in and near the soil in the vicinities of mines. The bacteria would also be genetically engineered so that they could be grown by untrained workers in the requisite large quantities and would not to be harmful after release into the environment.

Bacteria That Fluoresce in the presence of compounds emitted by buried mines would be grown on site, sprayed over a suspected minefield, then illuminated by ultraviolet light at night to observe glowing patches indicative of underlying mines.
In a typical envisioned application (see figure), the bacteria would be grown on site in drums, then sprayed over a suspected minefield. The operation would be timed to allow several hours after spraying for the bacteria to come into contact with the chemical(s) of interest in the soil and to obtain the visible signal. The bacteria would be detected at night by shining an ultraviolet light on the sprayed field and looking for the fluorescence, which would manifest itself as green glowing patches on the soil surface over buried mines. The glowing patches would typically be visible; however, it would be desirable to use an electronic photodetector to find weakly glowing patches.

Another novel method involves the use of swept acoustic resonance by means of an apparatus that would include a transmitting and a receiving ultrasonic transducer placed in contact with the ground. This apparatus would be used in conjunction with a previously developed metal detector, which would be used to locate buried metal objects that may or may not be the metal components of landmines (most landmines contain at least some metal). The ultrasonic transducers would be placed on the surface at a small safe distance away from a location indicated by the metal detector, so that the surface projection of the detected metal object would lie between the transducers. The transmitting transducer would emit a train of acoustic pulses shaped to have energy concentrated in the frequency range of 5 to 50 kHz. The structural features of landmines have mechanical resonances in this frequency range. The output of the receiving transducer would be processed to develop a transfer function that would exhibit a measurable characteristic in the presence of resonances associated with a landmine.

A third novel method involves exploitation nuclear quadrupole resonance (NQR). The basic idea is to excite and detect NQR by means of a lowpower signal at a suitable radio frequency applied to a small induction coil placed near a suspected mine. Previous research has already shown that some explosive compounds have NQR signatures useful for detection: for example, trinitrotoluene (TNT) exhibits a response distributed over 12 spectral lines, of which the ones most appropriate for detection of mines lie in the frequency range from 836 to 870 kHz.

A fourth novel method involves semiotic multi-sensor data fusion. The basic idea is to apply semiotic concepts (in contradistinction to older feature-vector-based pattern-recognition methods) to make the most effective use of data from multiple sensors in order to achieve both a high probability of detection and a low falsealarm rate. The approach to be taken in developing this method is not as well defined as are the approaches being followed for the other three methods. However, the mere fact that dogs can detect mines consistently by means that we do not yet understand can be taken as evidence that mines have detectable features.

This work was done by S. W. Kercel, R. S. Burlage, D. R. Patek, and C. M. Smith of Oak Ridge National Laboratory and A. D. Hibbs and T. J. Rayner of Quantum Magnetics, Inc.