Fast Quantitative Microwave Imaging Based on Calibration Measurements

Abstract

This thesis contributes to the solution of the inverse electromagnetic (EM) scattering problems arising in microwave imaging. A calibration technique based on measurements of specific objects is proposed and a fast quantitative imaging method based on such measurements is developed. The calibration measurements are performed on two known objects: the reference object representing the scatterer-free measurement and the calibration object representing a small scatterer embedded in the reference object. The inversion method does not need analytical or numerical approximations of the forward model as those are replaced by the measurement-based model. It is particularly valuable in short-range imaging, where analytical models of the incident field do not exist while the fidelity of the simulation models is often inadequate. In this thesis, it is demonstrated that the implementation of the calibration technique in the sensitivity-based imaging improves both the imaging efficiency as well as the image quality. A quantitative imaging method is further developed based on the calibration measurements where a direct inversion in real space is employed. The electrical properties of dielectric objects are reconstructed using a resolvent kernel in the forward model, which is extracted from the calibration measurements. The experimentally determined resolvent kernel inherently includes the particulars of the measurement setup, including all transmitting and receiving antennas. The inversion is fast, allowing for quasi-real-time image reconstruction. The theoretical limitations of the fast quantitative imaging method have been investigated and its performance with noisy data has been examined. It is found that the proposed method has limitations which are more flexible than those of the linear Born model. The method is also robust to random noise. Both the calibration technique and the fast quantitative imaging method are validated through synthetic, simulation and/or experimental examples. The proposed concept of experimentally derived resolvent kernel in the forward model is general and may be valuable in other imaging modalities such as ultrasound, photonic imaging, electrical-impedance tomography, etc.