Application of a library based method to time domain electromagnetic data for the identification of unexploded ordnance

A simple library based algorithm for the identification of unexploded ordnance is tested on time domain electromagnetic data. A library of polarization tensors is generated from data acquired on a test stand over a collection of different UXO. The objective of the algorithm is to determine which target of the library is most likely to have produced an observed data anomaly. For each target in our library a non-linear inverse problem is solved for the position and orientation that minimizes the least-squares difference between the observed data anomaly and the data predicted from each target. This technique avoids direct inversion for polarization tensor, making it feasible for cases where sensor data quality may not be sufficient to support confident estimation of model parameters. For cases where the background soil response is significant, we also invert for the t−1 decay characteristic of viscous remnant magnetic soil. We present results of preliminary tests of the library technique to Geonics EM63 time domain electromagnetic data collected on a test plot seeded with UXO. These tests demonstrate an excellent ability to accurately identify isolated targets. Misidentification of single targets occur when data anomalies have low signal to noise ratios or when targets within the library have similar polarization tensors.