Model-based detection of buried objects.

Abstract

An approach to the detection of buried objects is to excite the ground with low-frequency, high-energy sound waves, which then excites a resonance in the buried object. The ensuing vibration causes a detectable signal on the surface of the ground, which can be detected using a laser doppler vibrometer (LDV). The original detection technique used a sliding bandpass filter to process the scattered LDV energy, providing an energy map of the area scanned by the LDV, which indicates the location of the object. The performance of detection is often limited by speckle noise, a type of noise arising from the coherent nature of the laser beam. A more recent technique utilizes an autoregressive model of this noise. This leads to an inverse filter that whitens the noise. Upon the appearance of any target data in the signal, a whiteness test indicates a detection. This approach has demonstrated improvement over the bandpass filter approach. This paper demonstrates a further improvement by augmenting the prewhitener with a model of the mine itself. This provides significant improvement by both enhancing the mine signal and improving the detection performance. Experimental results are shown.