Fusion of anomaly algorithm decision maps and spectrum features for detecting buried explosive Hazards in forward looking infrared imagery

Remediation of the threat of explosive hazards is an extremely important goal. Such hazards are responsible for an unacceptable number of deaths and injuries to civilians as well as soldiers throughout the world. In this article, we put forth a new method for aggregating image space anomaly algorithm decisions across time (multi-look) as well as across disparate algorithms in Universal Transverse Mercator (UTM) space for forward looking vehicle mounted (FL) long-wave infrared (LWIR) imagery. We also explore the utility of fast Fourier transform (FFT) spectrum features, which were previously used for FL ground penetrating radar (FLGPR), on aggregated UTM anomaly algorithm decision (UTMAAD) maps. On a final note, we also discuss modifications to our pre-screener, an ensemble of trainable size contrast filters, for UTMAAD maps. Targets not detected at the moment are also not found by a human under visual inspection. Preliminary lane-based cross validation (CV) experiments are reported using field data measurements from a U.S. Army test site.