Enhanced landmine discrimination using the multi-axis electromagnetic induction response of a buried finitely conducting object

In the past, attempts to use low-frequency electromagnetic induction (EMI) techniques for military and humanitarian demining efforts have met with limited success due in large part to the many false alarms generated by metallic clutter. Currently, efforts are underway to exploit the quasi-magnetostatic response characteristics of metallic objects in order to discriminate between mines and clutter. A novel three-dimensional dipole model is used to represent the three cardinal induction response modes of a buried conducting object. The three cardinal modes of the buried object are excited and measured by orienting the sensor coils such that the excitation magnetic field is aligned along the cardinal axes of the target. The exact spatial locations needed to excite all modes are determined by observing polarizability plots of the transmitter-receiver coil pair. The three-dimensional dipole model provides three characteristic decay rates and three signal energy values for a total of six characteristic parameters that describe the response of a buried object. The majority of mines are cylindrical in nature, so the six characteristic parameters from the cardinal Cartesian modes are manipulated into five independent parameters describing the cylindrical modes of the target. The five independent parameters are the z-axis decay rate, z-axis energy, radial-axis decay rate, radial-axis energy, and a symmetry parameter. In order to determine if the source of the five observed parameters is a mine or a piece of clutter, a binary hypothesis-testing problem is formed. The observations are treated as random variables with known probability density functions (pdf)s. The pdfs for the observations under both hypothesis are estimated from known measured data. A generalized likelihood ratio test (GLRT) is formed to optimally decide between the two hypotheses. The GLRT is tested on unknown data collected at a blind test grid. Alternative forms of the GLRT are derived to operate on different subsets of the five observed parameters. The evaluation of the GLRTs is presented in the form of a receiver operator characteristic (ROC) curves that shows the probability of detection versus probability of false alarm for a range of threshold values. The ROCs show that improved clutter rejection is achieved as more independent parameters are used to characterize the buried objects.