الفهرس | Only 14 pages are availabe for public view |
Abstract In this thesis, some new techniques for the blind inversion of inhomogeneous media with arbitrary profiles are developed. The inversion process yields to the medium characteristics as a function of actual depths, such as the permittivity profiles (for lossless dielectric media), the permittivity and conductivity profiles (for lossy media) and the plasma and collision frequencies (ionospheric layers). The inversion is done blindly without aperiori knowledge about the profiles being reconstructed. The equivalence between the wave propagation through inhomogeneous ;,,1.Ub!’ structure and the signal transmission through IIR filters. The transfer function of the o equivalent filter contains all the information abut this structure including multiple reflections and reverberatory effects. In the case of lossless media, an impulsive plane wave is applied and the direct problem is solved using the IIR representation of the structure. Using the Dynamic Predictive Deconvolution (DPD) technique, the inversion process has been carried out with a great deal of accuracy in the presence of relatively high levels of noise. Both abrupt (discrete) and smooth (continuous) profiles are considered with low and high permittivity values. In the case of lossy non-dispersive media, a narrow band carrier modulated Gaussian pulse is used as a probing signal. Two techniques have been tested in this case namely the IIR representation of the structure and the layer by layer representation. Using both techniques, the reconstruction process has been shown to give almost the same results. Although the IIR representation is more general, we found that it is more complicated compared with the other. Also, in this case, accurate inversion of permittivity and conductivity profiles (abrupt and smooth) has been demonstrated. These inversion techniques have also been extended to dispersive media. As an example, the inversion of an ionospheric layer was carried out and the plasma and collision frequencies profiles have been successfully reconstructed. In all of the considered techniques and examples, the number of the structure layers and their physical depths are accurately determined using the correlation characteristics of the reflection data and the sampling time. Also, the effect of noise (or measurement errors) has been considered. Good noise performance has been observed in these techniques. |