Airport unexploded ordnances identification based on artificial neural network and fuzzy support vector machines

Considering the instability of airport unexploded ordnances(UXO) and the complexity of environment, the theory of artificial neural network(ANN)-fuzzy support vector machines (FSVMs) is presented to penetrate UXO. Different from the traditional target identification methods, the proposed approach uses the characteristics of ground penetrating radar target data analyzed by using the principal component analysis (PCA) technique. Considered many coterminous characteristics data of the targets, they are classified with a combination of support vector classifiers (SVCs) and feed forward neural networks (FFNNs). The risk membership to each input points is confirmed on the base of processing input data, and then is leaded into the reasoning process of the decision function. The results of UXO show that the proposed approach gives accurate results in terms of the estimated UXO identification.