Application and extension of texture feature coding methods to anti-tank landmine detection
Recent advances in ground penetrating radar (GPR) fabrication and related signal processing have yielded robust performance on government sponsored blind tests of anti-tank landmine detection capabilities on test lanes. Recent data collections with the NIITEK GPR system have focused on more difficult "off-lane" soil conditions that typically contain higher levels of sub-surface GPR anomalies and provide more difficult tests of anti-tank target detection capabilities. Our recent research in this field has focused on the application of advanced signal processing techniques to target/clutter discrimination at pre-screener-flagged locations of interest. In this work we discuss the applications and extensions of a texture feature coding method (TFCM) for landmine detection in off-lane soils. First we consider application of the TFCM technique to target detection in 2-D GPR data slices. We also consider application of the TFCM to "tiled" images containing multiple instantiations of a target response. Finally we consider a 3-D extension of the TFCM and apply our extension to target detection in 3-D time-domain GPR data. Our results indicate performance increases for TFCM-based processing of prescreener generated alarms, with the most robust performance increases resulting from application of our 3-D TFCM extension.
