In this paper, a new deep learning approach is proposed to solve the electromagnetic inverse scattering (EMIS) problems. The conventional methods for solving inverse problems face various challenges including strong ill-conditions, high contrast, expensive computation cost and unavoidable intrinsic nonlinearity. To overcome these issues, we propose a new two-step machine learning based approach. In the first step, a complex-valued deep convolutional neural network (DConvNet) is employed to retrieve initial contrasts (permittivities) of dielectric scatterers from measured scattering data. In the second step, the previously obtained contrasts are input into a complex-valued deep residual convolutional neural network (DRCNN) to refine the reconstruction of images. Consequently, the EMIS problem can be solved with much higher accuracy even for high-contrast objects. Numerical examples have demonstrated the capability of the newly proposed method with the improved accuracy. The proposed deep learning approach for EMIS problem serves a new path for realizing real-time quantitative microwave imaging for high-contrast objects.

This paper proposes a novel deep learning (DL) approach to resolve the electromagnetic inverse scattering (EMIS) problems. The conventional approaches of resolving EMIS problems encounter assorted difficulties, such as high contrast, high computational cost, inevitable intrinsic nonlinearity and strong ill-posedness. To surmount these difficulties, a novel DL approach is proposed based on a novel complex-valued deep fully convolutional neural network (DConvNet) structure. The proposed complex-valued deep learning model for solving EMIS problems composes of an encoder network and its corresponding decoder network, followed by a final pixel-wise regression layer. The complex-valued encoder network extracts feature fragments from received scattered field data, while the role of the complex-valued decoder network is mapping the feature fragments to retrieve the final contrasts (permittivities) of dielectric scatterers. Hence, the proposed deep learning model functions as an “heterogeneous” transformation process, where measured scattering field data is converted into the corresponding contrasts of scatterers. As a consequence, the EMIS problem could be resolved accurately even for extremely high-contrast targets. Numerical benchmarks have illustrated the feasibility and accuracy of our proposed approach. The proposed approach opens a novel path for the deep learning based real-time quantitative microwave imaging for high-contrast scatterers.

In this paper, a novel absorbing boundary condition (ABC) computation method for Finite-Difference Time-Domain (FDTD) is proposed based on the machine learning approach. The hyperbolic tangent basis function (HTBF) neural network is introduced to replace traditional perfectly matched layer (PML) ABC during the FDTD solving process. The field data on the interface of conventional PML are employed to train HTBF based PML model. Compared to the conventional approach, the novel method greatly decreases the size of computation domain and the computation complexity of FDTD because the new model only involves the one-cell boundary layer. Numerical examples are provided to benchmark the performance of the proposed method. The results demonstrate that the newly proposed method could replace conventional PML and could be integrated into FDTD solving process with satisfactory accuracy and compatibility to FDTD. According to our knowledge, this proposed model combined ANN model is an unreported new approach based on machine learning based for FDTD.

Deep learning is applied to assist the joint inversion for audio-magnetotelluric and seismic travel time data. More specifically, deep residual convolutional neural networks (DRCNNs) are designed to learn both structural similarity and resistivity-velocity relationships according to prior knowledge. During the inversion, the unknown resistivity and velocity are updated alternatingly with the Gauss-Newton method, based on the reference model generated by the trained DRCNNs. The workflow of this joint inversion scheme and the design of the DRCNNs are explained in detail. Compared with describing the resistivity-velocity relationship using empirical equations, this method can avoid the necessity in modeling the correlations in rigorous mathematical forms and extract more hidden prior information embedded in the training set, meanwhile preserving the structural similarity between different inverted models. Numerical tests show that the inverted resistivity and velocity have similar profiles, and their relationship can be kept consistent with the prior joint distribution. Furthermore, the convergence is faster, and final data misfits can be lower than separate inversion.

The electromagnetic inverse scattering (EMIS) problem is solved by a novel two-step deep learning approach in this paper. The newly proposed two-step deep learning approach does not only predict the multi-frequency EM scattered field, but also overcomes the limitation of the conventional methods for solving EMIS problems, such as expensive computational cost, strong ill-conditions and invalidity on high contrast. In the first step, the complex-valued deep residual convolutional neural network (DRCNN) is utilized to predict of multi-frequency EM scattered field only using single-frequency EM scattered field information. Based on a new complex-valued deep convolutional encoder-decoder (DCED) structure, the second step utilizes the obtained multi-frequency EM scattered field ‘images’ to realize reconstruction of the target scatterers. In such a manner, the proposed approach can solve EMIS problem accurately and efficiently even for inhomogeneous and high-contrast scatterers. The training of the proposed two deep learning models is based on the simple synthetic dataset. Numerical examples based on various dielectric objects are given to demonstrate the accuracy and performance of the newly proposed approach. The proposed deep learning based method opens a new path for handling the real-time quantitative microwave imaging.

A low-profile, wideband and circularly polarized (CP) magneto-electric (ME) antenna with high gain and end-fire radiation beam is proposed in this paper. It consists of a magnetic dipole (M-dipole) and two pairs of printed crossed-strips to produce the vertically and horizontally polarized E-field components, respectively. With the appropriate combination, the proposed ME antenna exhibits an inherent 90°-phase difference at endfire for these two field components. CP end-fire radiation, therefore, can be achieved. Two techniques, namely, using the L-probe coupled feeding structure to excite the M-dipole operating at both the basic (0.5»-) and higher (1.5»-) modes for the dual-mode resonances and using two pairs of printed strips with different dimensions for the dual-frequency resonances, are introduced to realize the wideband operation. Measurement results show that the wide impedance bandwidth (IMBW) of 2.40-4.24 GHz (1.84 GHz, 55.4%) and 3-dB axial-ratio bandwidth (ARBW) of 2.36-3.86 GHz (1.5 GHz, 48.2%) are achieved. Besides, the stable end-fire radiation patterns with a peak gain of 7.57 dBi are observed over the operating band. And the bandwidth for 3-dB gain variation is 2.34-3.97 GHz. Hence, the proposed antenna with its main beam in parallel with the substrate shows the potential for some point-to-point communications such as the radio-frequency identification (RFID) system.