Reseach On Target Imaging And Inversion Scheme In Heterogeneous Medium

Target inversion and imaging in heterogeneous medium exists in many fields. It has strong demand and great value to investigate the modeling of heterogeneous medium and imaging method of the target inside. This dissertation starts from the propagating facts of wave in heterogeneous medium. Based on the application background of buried object imaging and biotissue imaging, the two major scientific questions are extracted as modeling the medium and model-based target imaging. For the modeling problem, the iterative time reversal mirror(TRM) method and the active adjoint modeling(AAM) method are proposed. For the imaging problem, besides the TRM imaging method is well ultilized, the model-based compressive sensing(CS) method is proposed to reduce the data redundancy during scanning and imaging targets with spatial sparsity.Aiming to model the boundary of layered medium, an iterative TRM modeling method is proposed and tuned for underground object imaging and steel frame imaging in concrete. The geometric model of the buried object imaging problem is established. The boundary of ground medium is spatial discreted into scattering sectors. The proposed iterative TRM method is used to modeling the boundary and also to image the target beneath it. The mathematic theory of imaging multiple scatterers by TRM is explained by MUltiple SIgnal Classification(MUSIC). Numerical simulations based on FDTD are performed to evaluate the modeling and imaging performance. Furthermore, multi-pulses transmitted in HH/VV polarization are used to perform the iterative TRM imaging of steel frame in concrete wall.Aiming to modeling the medium with structural heterogeneity, the adjoint modeling method is proposed and tuned for MITAT on breast tumor. Starts from the general notation form of differential equation, the derivative of misfit function on model parameters is calculated by adjoint method, with the example of 2-D scale wave equation. The iterative conjugate gradient optimization method is used to minimize the misfit function and output the modeling results. In MITAT, the active adjoint modeling(AAM) method are introduced. Numerical simulations on checkerboard pattern model is performed for verification and robustness analyze on noise. To further evaluate the proposed AAM method, an anatomically realistic breast(ARB) model based on MRI data is made and used as the target model of AAM. The modeling result by AAM significantly improved the TRM imaging quality of tumors inside the ARB model.Aiming to reduce the data acquisition complexity of model-based imaging approach in MITAT, i.e. TRM imaging, a CS-MITAT method with model-based dictionary is proposed. It well utilized the spatial sparsity of early stage breast tumors. Based on the physical characteristics of MITA signals, the dictionary of tumor signal is made from a priori or reconstructed heterogeneous breast model. Random spatial observation matrix is defined to perform the compressing procedure of CS. The state of the art GPSR method is applied to recover the distribution information of tumors. Numerical simulation and real biotissue experiments on an MITAT prototype system both show that the CS-MITAT significantly reduced the data acquisition time while maintaining the same tumor imaging quality.A thoroughly investigation on the modeling of heterogeneous medium and target imaging has been done in this dissertation, along with carefully chosen targeted research in this field. The two modeling methods and one data acquisition/imaging method is verified and evaluated by numerical simulations and experiments. Factors that could affect modeling and imaging performance are also discussed. This work has established a reference for further studies in heterogeneous medium related problems.