Autonomous UXO classification using fully polarimetric GPR data

This paper presents an automatic UXO classifcation systcm using neural network and fuzzy inference based on the clmiiication rules. These rules were based on the scattering behavior predicted from various canonical shapes related to common UXO and clutter items in actual UXO sites. In this paper, the classification rules were also verified and modified by the method of moment simulation. The rules were then implanted to an expert system consisting of two stages. The first-stage classifies objects into clutter (group-A and D), a horizontal linear objelct (group-B) and a vertical linear object (group-C) according to the spatial distribution of the Estimated Linear Factor (ELF) values through the neural network. Then second-stage discriminates UXO-LIKE targets among objects under groups B and C by fuzzy inference with quantitative variables. The classifcation performance of this automatic algorithm is comparable with or superior to that obtained from a trained expert. However, the automatic classification procedure does not require the involvement of the operator and assigns an unbiased quanititative confidence level associated with each classification. Classification error and inconsistency associated with fatigue, memory fading or complex features should be greatly reduced.