Landmine detection with a standoff acoustic/laser technique

Abstract

Landmines and mine-like traps are effective weapons that are difficult to detect and discriminate from a safe distance. The ability to detect landmines in their host environment at a distance and to discriminate them from other objects would be valuable for countering the landmine threat. This paper explores a standoff acoustic/laser technique to discriminate landmines from other forms of man-made objects (clutter) in an urban environment. A novel approach currently under investigation by MIT Lincoln Labs, University of Mississippi, and other groups employs a non-contact acoustic/laser technique to detect landmines from a safe standoff range. This technique uses a sound source to excite vibrations in targets with an acoustic wave. These vibrations are in turn measured remotely with a Laser Doppler Vibrometer (LDV). In this thesis, the vibration responses of landmine variants are measured, analyzed, and compared to those of common urban objects likely to be found on a landmine field or roadside. The Fourier Transform of the vibration of the target as measured by the LDV is used to generate a target vibration spectrum. Target vibration spectra in response to a sound source were experimentally measured for 59 trials, 28 of which were of simulated landmine variants and the remaining trials were of urban clutter objects. Using an algorithm adapted from a methodology for mass spectral analysis, parameters of the target signatures are estimated; then individual target signatures are classified using a Support Vector Machine (SVM) with a training set composed of parameters from the remaining members of the total population. The best results obtained from this methodology had a 71% probability of detection and a 3% false alarm rate corresponding to 20 of 28 of the simulated landmine variants correctly identified and a single clutter object misidentified as a landmine variant.