1 October 2024 in IEEE Transactions on Antennas and Propagation
In this study, we present a novel approach to address non-weak scattering problems by integrating deep learning into the Born series. Typically, the first order Born approximation is limited to cases where the contrast between the scatterer and background medium is exceptionally low. While higher order terms in the Born series can be used for higher contrasts, convergence issues may arise due to highly oscillatory factors in Green’s function. To overcome this limitation, we introduce a physics-based deep learning method inspired by the Born series, which effectively predicts the distribution of the complex electromagnetic field. The proposed series assures convergence when scattering occurs from high-contrast objects, due to the use of a learning-based forward operator. Exploiting the physics-based nature of our model, we adopt a simple convolutional neural network (CNN) architecture, requiring significantly fewer training data. Our results demonstrate very good generalization capabilities of the proposed approach, showcasing its ability to handle unseen background fields and profiles. We deem this innovative series as an extension of the Born series that can be effectively employed in highly nonlinear problems.
14 July 2014 in IEEE Transactions on Geoscience and Remote Sensing
The integral equation technique, in combination with the method of moments, is widely used for treating the electromagnetic scattering from periodic rough surfaces. This method, however, requires the computation of a kernel (a periodic Green's function) that consists of a slowly converging infinite series of conventional Green's functions. To accelerate the convergence of this series, different methods have been proposed in literature. In this paper, we use a complex image (CI) technique to find a closed-form expression for the periodic Green's function, which is then used to analyze the electromagnetic scattering from arbitrary periodic rough surfaces. The resulting CI Green's function consists of a finite number of terms corresponding to real and complex sources. Moreover, the CI Green's function in the Hankel form (evanescent waves) is used if the observation and source points are close. In the far-field region, the series in the plane-wave form (propagating Floquet modes) caused by the extraction of poles are sufficient and necessary. The results indicate that the presented CI Green's function leads to numerical efficiency yet having a rather high level of accuracy.
12 October 2020 in Journal of the Optical Society of America B
MKMohammad Ali Khosrovani-Moghaddam
Mohammad Ali Khosrovani-Moghaddam
Motivated by surface plasmon polariton waveguides in the optical regime, spoof surface plasmon (SSP) waveguides have received a lot of attention in terahertz and millimeter wave frequencies. Most research on these kinds of waveguides is numerical. However, some limited analytical work can be seen in the literature. In this paper, one type of SSP waveguide that is composed of a rectangular corrugation with finite lateral width on the ground is considered, and an analytical method, which is inspired by Marcatili’s method, is proposed in order to calculate the dispersion curve of the first mode. The results of this analytical method and a numerical commercial eigenmode solver are compared. The accuracy of this method by varying different geometrical parameters is shown. Finally, it is shown that this method works well in predicting the electromagnetic fields of the first mode.
13 June 2023 in IEEE Transactions on Antennas and Propagation
This work presents a novel closed-form expression based on the Method of Moments (MoM) for solving the scattering problem in piecewise homogeneous media. Using point-matching with pulse basis functions in MoM, we derive an expression represents the scattered field as an explicit function of the permittivity value of each homogeneous part. The coefficient matrix depends on the permittivity values in the MoM solution to the scattering problem. Therefore, it requires matrix inversion whenever these values change slightly. Here, we propose a new expression where the matrix coefficients are independent of the permittivity values of each homogeneous part. Accordingly, when the geometry does not change, we propose a semi-explicit procedure for updating the solution with less computational complexity. Although the proposed method can be used for all 2D and 3D problems, the formulation details are implemented for 2D problems, TM incidence. Based on this explicit form, we derive some simple approximate expressions for three specific cases of permittivity perturbations. Our proposed solution is appropriate for microwave imaging, particularly when a priori information is utilized.
27 February 2024 in IEEE Transactions on Antennas and Propagation
In microwave imaging (MI), utilizing prior boundary information to enhance the accuracy of retrieved dielectric properties holds great promise. However, the precise registration of prior boundaries with the actual boundaries of the imaged object is crucial, as misalignment can result in significant inaccuracies in retrieved values. To address this issue, we present a novel approach that formulates the problem of aligning the prior and actual boundaries as a shape optimization problem. The proposed method employs the free-form deformation (FFD) technique to parameterize the shape optimization problem. Furthermore, we introduce a new method for efficient computation of the Jacobian matrix by analyzing the sensitivity of the scattered field with respect to the location of control points in the FFD. Through simulations of microwave imaging of the lower extremity of the human body, we demonstrate the effectiveness of our proposed approach. Our results indicate that the proposed approach reduces the average Hausdorff distance between the prior and actual boundaries by approximately 80%, leading to a substantial enhancement in the accuracy of retrieved dielectric properties.