19 January 2009 in IEEE Transactions on Microwave Theory and Techniques
A new microwave imaging technique is proposed for reconstruction of 2-D complex permittivity profiles in dielectric samples located in a waveguide system. The spatial distributions of the dielectric constant and the loss factor are approximated by continuous functions whose functional parameters are determined using a neural network technique backed by full-wave finite-difference time-domain analysis. The profiles are reconstructed from measurements of reflection and transmission characteristics obtained with the tested sample at different locations. Operational capabilities of the technique are illustrated through a series of computational experiments for rectangular and cylindrical samples at two (original and 90deg -rotated) positions. The results demonstrate excellent agreement between the reconstructed and actual profiles approximated by linear, quadratic, and Gaussian functions: the average relative errors do not exceed 0.4%, 2.2%, and 4.8%, respectively. Finally, the assumption of functional approximation, uniqueness of the reconstruction, and prospects of practical use of the technique are thoroughly discussed.
12 September 2008 in IEEE Microwave and Wireless Components Letters
A new 3-D microwave imaging technique is proposed for the detection of a position and size of a spherical object in a dielectric sample located in a waveguide system. The reconstruction is based on a radial basis function neural network inversion, backed by finite difference time domain analysis and requires only elementary measurements of complex reflection and transmission coefficients. Functionality of the technique is illustrated by computational experiments in reconstructing the parameters of a glass sphere and an air bubble in a rectangular Teflon block. It is shown that, at 915 MHz, the spheres of not less than 15 mm diameter are reconstructed with the average errors of 0.9%-2.2%.
1 March 2003 in Optical and Quantum Electronics