Simulation study of SVM-based approach in extracting parameters for electromagnetic detection of buried object

In this paper, a new method for extracting the parameters of buried object is proposed. The center position and dielectric properties of 2-D buried object are estimated by means of a regression technique based on support vector machine (SVM). The proposed method, after a proper training procedure, is able to reconstruct the center position and dielectric properties of a buried object inside a given investigation domain. Numerical simulation results indicate that SVM-based approach shows higher accuracy than the back-propagation neural networks (BPNN) algorithm.