Imaging of a micro-structure: binary contrast source inversion and convolutional neural networks

Peipei Ran (1) , Yingying Qin (2, 1) , Dominique Lesselier (1) , Mohammed Serhir (3)

1 L2S - Laboratoire des signaux et systèmes
2 SATIE - Systèmes et Applications des Technologies de l'Information et de l'Energie
3 GeePs - Laboratoire Génie électrique et électronique de Paris

Time-harmonic transverse-magnetic electromagnetic imaging of a grid-like, finite set of circular cylindrical dielectric rods placed in air with sub-wavelength distances between adjacent rods and sub-wavelength diameters is considered. It is shown how one can be achieving super-resolution in the resulting micro-structure, reaching beyond the Rayleigh criterion, using a binary-specialized form of teh contrast source inversion method and another method based on convolutional neural networks. illustrations of both methods are proposed from comprehensive numerical simulations in a number of configurations of interest, this seen in terms of organization of the micro-structure itself, discrete frequencies of observation, data acquisition and data noise. Experimental validations are proposed as well from experiments carried outt in an anechoic chamber.

Keywords: microstructure; inversion; imaging; binary priors; contrast source inversion; convolutional neural networks

Domains: Electromagnetism; Signal and Image processing; Mathematical Physics [math-ph]

Peipei Ran, Yingying Qin, Dominique Lesselier, Mohammed Serhir. Imaging of a micro-structure: binary contrast source inversion and convolutional neural networks. XXXIII General Assembly and Scientific Symposium (GASS) of the International Union of Radio Science, Aug 2020, Rome, Italy. pp.B06-02. ⟨hal-02906373⟩