Search

In this article, a unified learning-based approach is introduced to solve inverse scattering problems (ISPs) with mixed boundary conditions (BCs). The scattering behavior of hybrid dielectric and perfect electric conductors (PEC) scatterers is modeled by the T-matrix method. A rough image of the zero-order T-matrix coefficients for unknown scatterers is first reconstructed by the backpropagation (BP) method, which is then refined by an attention-assisted pix2pix generative adversarial network (GAN). The spatial attention mechanism is utilized to enforce the generator network to learn salient features of the unknown scatterers instead of the background. The adversarial training of the generator and the discriminator further enables the reconstructed image to be constrained by high-level features of reference scatterers. Numerical tests on both synthetic and experimental data verify the superior performance of the proposed method for ISP reconstructions with hybrid scatterers. It effectively expands the application scope of learning-based ISP methods to reconstruct scatterers without knowing the BCs of scatterers in advance. [ABSTRACT FROM AUTHOR]

In the mixed boundary inverse scattering problem (ISP), conducting and dielectric scatterers coexist in the same region, which challenges the present quantitative inverse scattering methods. Moreover, to ensure the incident waves penetrating the lossy or high-contrast objects, lower wavelength is applied in most inverse scattering applications. Therefore, methods with wavelength or subwavelength resolution are required for the mixed boundary ISP. In this article, we devise a quantitative inversion scheme alternately updating the contrast of dielectric scatterers and the T-matrix of conducting scatterers. The proposed alternate parameter updating method (APUM) avoids the reconstruction deterioration from both the large imaginary parts of conducting contrasts and the limited expansion order of the T-matrix. Then, we further improve the resolution of the APUM by optimizing the incident fields, which is also a regularization strategy. In particular, we design superoscillatory incident fields to quantitatively converse the high spatial spectrum of objects into low spectrum contrast sources, which can retain the high-frequency information of objects in the low-passband of the Green function. The results with synthetic data and single-frequency Fresnel experimental data verify the effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]