Research On Electromagnetic Inverse Scattering Enhanced Imaging Algorithm

The electromagnetic inverse scattering technique breaks through the limit of electromagnetic diffraction and reconstructs target parameters in the detection area,achieving high-resolution imaging and widely applied in civilian and military fields.In practical applications,the imaging resolution is often affected by the nonlinearity and illposedness problems in inverse problems.Therefore,how to effectively solve these problems,invert target parameters quickly,and further improve the imaging quality is the research hotspot of electromagnetic inverse scattering.This thesis focuses on the above-mentioned issues and proposes corresponding solutions to achieve inverse scattering-enhanced imaging.Firstly,this thesis establishes an electromagnetic forward scattering model based on a twodimensional scene and deeply studies the application of the conjugate gradient-fast Fourier transform(CG-FFT)method in obtaining forward scattering data.Based on the integral equation,this thesis analyzes in detail the distillation process of the forward scattering model using the method of moment(MOM)and the implementation of the CG-FFT iteration algorithm,and verifies the effectiveness of this method in simulation experiments.In the actual project,the MOM and CG-FFT methods are used to simulate and calculate the scattering data of actual scenes.This thesis uses a precise calibration method to fit the measured data,and compensates for the offset error of the simulated field,improving the accuracy and reliability of the experimental data.Secondly,this thesis proposes a subspace-based optimization method combining adaptive total variation regularization technology(Adaptive Total Variation Regularization Subspacebased Optimization method,ATVR-SOM)to address the ill-posedness problem in electromagnetic inverse scattering imaging.The method distinguishes edges and slopes based on the gradient information of the image,and automatically selects better models near the edge and smoother models further away from the edge.This model not only reduces the possibility of misjudging noise as an edge but also mitigates the "staircase effect" introduced by the TV regularization factor,making the reconstructed image edge information clearer.This thesis verifies the effectiveness of the ATVR-SOM method through numerical simulations and experimental tests.Finally,this thesis proposes a comparison source inversion method based on a hybrid conjugate gradient to address the slow or even non-convergence problem when iteratively inverting actual scattered data with noise using nonlinear iterative algorithms.This method corrects the original conjugate parameters through a convex combination to ensure that the gradient information determining the contrast at each iteration can achieve optimal convergence.The experimental results show that this method has faster convergence speed and better stability when dealing with noisy data,and the reconstructed image is clearer.