An artificial neural network application for buried target classification

Design of an electromagnetic target classifier is demonstrated in this study. The surface equivalence principle and the method of moment are used to calculate the electric field scattered from buried cylindrical target. A database is obtained by the scattered fields from two cylindrical targets at certain frequencies. While a portion of the database is used to train the network, the rest is used to test the performance of the neural network. Radial basis function network is proposed for target classification. This target classification process is repeated by changing the surface roughness and electrical parameters of the medium in which the target is buried, incident angle of the incident wave and cross section of the cylindrical target. Effects of these parameters to the target classification are investigated.