In our previous work, we have shown that the detectability of landmines can
be improved dramatically by the careful application of signal detection th
eory to time-domain electromagnetic induction (EMI) data using a purely sta
tistical approach. In this paper, classification of various metallic land-m
ine-like targets via signal detection theory is investigated using a protot
ype wideband frequency-domain EMI sensor. An algorithm that incorporates bo
th a theoretical model of the response of such a sensor and the uncertainti
es regarding the target/sensor orientation is developed. This allows the al
gorithms to be trained without an extensive data collection. The performanc
e of this approach is evaluated using both simulated and experimental data.
The results show that this approach affords substantial classification per
formance gains over a standard approach, which utilizes the signature obtai
ned when the sensor is centered over the target and located at the mean exp
ected target/sensor distance, and thus ignores the uncertainties inherent i
n the problem. On the average, a 60% improvement is obtained.