Target Shape Classification Using Electromagnetic Induction Sensor Data

Classification based on shape is central to the problem of discriminating between unexploded ordnance (UXO) and clutter. Most UXO fit a specific profile: they are long and slender with typical length-to-diameter aspect ratios of four or five. Most clutter items, on the other hand, do not fit this profile. Using electromagnetic induction (EMI) sensor data, we present a model-based estimation procedure to determine whether or not a target is likely to be a UXO item. The model relies on exploiting the dependence of the induced field on target size, shape and orientation. Carefully collected measurements using a modified EM61 from the Naval Research Laboratory (NRL) Multi-sensor Towed Array Detection System (MTADS) are used to establish this dependency.