Landmine classification based on high-resolution temporal-spatial GPR template

This paper presents an automatic landmine clasisification approach using a high spatial and temporal resolution data based on 2D templates in a typical GPR format. Suclh a higher resolution data were obtained from a slightly elevated dielectric rod probe [l] This arrangement in conjunction with broad frequency band and multiple-position from a linear scam over a target generate a two-dimensional (2D) scattering temlplate. The 2D scattering templates were trained into the neural networks (NN) to perform the landmine classification autalmatically by comparing the template with test target and generated the type of a mine and confidence level as a result. The autanmatic classification algorithm was tested with blind target sets. The performance of the automatic classification system was found to be very promising compared with that achieved by a trained expert..