Applying Convolutional Neural Networks for the Source Reconstruction

This paper proposes a novel source reconstruction method (SRM) based on the convolutional neural network algorithm. The conventional SRM method usually requires the scattered field data oversampled compared to that of target object grids. To achieve higher accuracy, the conventional SRM numerical system is highly singular. To overcome these difficulties, we model the equivalent source reconstruction process using the machine learning. The equivalent sources of the target are constructed by a convolutional neural networks (ConvNets). It allows us to employ less scattered field samples or radar cross section (RCS) data. And the ill-conditioned numerical system is effectively avoided. Numerical examples are provided to demonstrate the validity and accuracy of the proposed approach. Comparison with the traditional NN is also benchmarked. We further expand the proposed method into the direction of arrival (DOA) estimation to demonstrate the generality of the proposed procedure.

[1]  N. S. Hoang,et al.  On node distributions for interpolation and spectral methods , 2013, Math. Comput..

[2]  Xian Liang Wu,et al.  Compressive Sensing for Fast Analysis of Wide-Angle Monostatic Scattering Problems , 2011, IEEE Antennas and Wireless Propagation Letters.

[3]  He Ming Yao,et al.  Machine learning based method of moments (ML-MoM) , 2017, 2017 IEEE International Symposium on Antennas and Propagation & USNC/URSI National Radio Science Meeting.

[4]  Phil Kim,et al.  MATLAB Deep Learning , 2017, Apress.

[5]  Alexander G. Yarovoy,et al.  Linearized 3-D Electromagnetic Contrast Source Inversion and Its Applications to Half-Space Configurations , 2017, IEEE Transactions on Geoscience and Remote Sensing.

[6]  Tapan K. Sarkar,et al.  AN ALTERNATIVE SPHERICAL NEAR FIELD TO FAR FIELD TRANSFORMATION , 1997 .

[7]  Jason Cong,et al.  Minimizing Computation in Convolutional Neural Networks , 2014, ICANN.

[8]  Jonathan Blackledge.  Introduction to imaging from scattered fields , 2015 .

[9]  R. G. Ayestaran,et al.  Near Field to Far Field Transformation Using Neural Networks and Source Reconstruction , 2006 .

[10]  Y. Alvarez,et al.  On the Comparison Between the Spherical Wave Expansion and the Sources Reconstruction Method , 2008, IEEE Transactions on Antennas and Propagation.

[11]  Xiaofei Zhang,et al.  Direction of Departure (DOD) and Direction of Arrival (DOA) Estimation in MIMO Radar with Reduced-Dimension MUSIC , 2010, IEEE Communications Letters.

[12]  L. J. Jiang,et al.  The Far Field Transformation for the Antenna Modeling Based on Spherical Electric Field Measurements , 2012 .

[13]  Berkman Sahiner,et al.  Classification of mass and normal breast tissue: a convolution neural network classifier with spatial domain and texture images , 1996, IEEE Trans. Medical Imaging.

[14]  Kurt Hornik,et al.  Multilayer feedforward networks are universal approximators , 1989, Neural Networks.

[15]  Matteo Pastorino,et al.  A Gibbs random field-based active electromagnetic method for noninvasive diagnostics in biomedical applications , 1995 .

[16]  M. D'Urso,et al.  THE CONTRAST SOURCE-EXTENDED BORN MODEL FOR 2D SUBSURFACE SCATTERING PROBLEMS , 2009 .

[17]  Mats Gustafsson,et al.  Reconstruction of equivalent currents using a near-field data transformation - with radome applications , 2004 .

[18]  Huan Huan Zhang,et al.  Coherent Processing and Superresolution Technique of Multi-Band Radar Data Based on Fast Sparse Bayesian Learning Algorithm , 2014, IEEE Transactions on Antennas and Propagation.

[19]  P. Hansen.  Discrete Inverse Problems: Insight and Algorithms , 2010 .

[20]  P. M. van den Berg,et al.  Two‐dimensional location and shape reconstruction , 1994 .

[21]  G. Vecchi,et al.  Field And Source Equivalence In Source Reconstruction On 3D Surfaces , 2010 .

[22]  Geoffrey E. Hinton,et al.  ImageNet classification with deep convolutional neural networks , 2012, Commun. ACM.

[23]  Yonina C. Eldar,et al.  Direction of Arrival Estimation Using Co-Prime Arrays: A Super Resolution Viewpoint , 2013, IEEE Transactions on Signal Processing.

[24]  Xiaoou Tang,et al.  Image Super-Resolution Using Deep Convolutional Networks , 2014, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[25]  A. Ng.  Feature selection, L1 vs. L2 regularization, and rotational invariance , 2004, Twenty-first international conference on Machine learning - ICML '04.

[26]  Yijun Yu,et al.  Three-Dimensional Bayesian Inversion With Application to Subsurface Sensing , 2007, IEEE Transactions on Geoscience and Remote Sensing.

[27]  L.F. Herran,et al.  High-accuracy neural-network-based array synthesis including element coupling , 2006, IEEE Antennas and Wireless Propagation Letters.

[28]  R. G. Ayestaran,et al.  Neural networks and equivalent source reconstruction for real antenna array synthesis , 2003 .

[29]  Ladislau Matekovits,et al.  A methodology for translating non-radiating sources in design parameters of cloaking devices , 2016, 2016 IEEE International Symposium on Antennas and Propagation (APSURSI).

[30]  W. Chew,et al.  Reconstruction of two-dimensional permittivity distribution using the distorted Born iterative method. , 1990, IEEE transactions on medical imaging.

[31]  Ji Wu,et al.  Study on a Poisson's equation solver based on deep learning technique , 2017, 2017 IEEE Electrical Design of Advanced Packaging and Systems Symposium (EDAPS).

[32]  Yuanyuan Zhang,et al.  Adaptive Convolutional Neural Network and Its Application in Face Recognition , 2016, Neural Processing Letters.

[33]  Fan Yang,et al.  Study on a 3D Possion's Equation Slover Based on Deep Learning Technique , 2018, 2018 IEEE International Conference on Computational Electromagnetics (ICCEM).