Matched neural filters for EMI based mine detection

Remedial mine detection and the detection of unexploded ordnance (UXO) have become very important for humanitarian reasons. This paper addresses mine detection using commonly used Electromagnetic Induction sensors. We propose and evaluate two neural network approaches to mine detection which provide a robust non-parametric technique, based on training the networks using data from a previously calibrated portion of the mine eld, or from a similar mine eld. In the rst approach, we combine a novel statistic, the S-Statistic (which is a real valued variable related to the relative energy di erence measured around a point in the mine eld) with the recently published -Technique [1] in a Random Neural Network (RNN) [2, 3, 4] design. In the second approach, a RNN is trained using a 3 3 block measurement window, and then applied as a post-processor for the -Technique. This RNN has an unconventional feedforward structure which realizes a matched lter to discriminate between non-mine patterns and mines. Experimental results for both approaches show that the RNN reduces false alarms substantially over the -Technique and the energy detector.