Retrieving missing elements of a 2-D micro-structure: joint-sparsity inversion and convolutional neural networks

Abstract : A finite number of infinitely long, regularly-distributed dielectric rods is illuminated from the outside by a set of time-harmonic ideal electric line sources and the fields scattered in each experiment is collected accordingly. Some of the rods are missing, associated to a lacunary (damaged) micro-structure, here within the hypothesis of radii of rods and inter-rod distances small versus the wavelength of operation. To detect those missing rods, a joint-sparsity non-linearized inversion can be applied as well as convolutional neural networks, with pros and cons that may differ, provided that a proper direct model based on a rigorous cylindrical wave expansion in the former case and a standard Method of Moments in the latter case is available. Elements of those two drastically different solution approaches are illustrated from numerical simulations, and super-resolution confirmed in both cases.
Complete list of metadatas

https://hal-centralesupelec.archives-ouvertes.fr/hal-02074720
Contributor : Dominique Lesselier <>
Submitted on : Wednesday, March 20, 2019 - 7:38:47 PM
Last modification on : Wednesday, September 18, 2019 - 9:24:01 AM

Identifiers

  • HAL Id : hal-02074720, version 1

Citation

Peipei Ran, Yingying Qin, Dominique Lesselier, Mohammed Serhir. Retrieving missing elements of a 2-D micro-structure: joint-sparsity inversion and convolutional neural networks. 9th International Conference on New Computational Methods for Inverse Problems, May 2019, Cachan, France. ⟨hal-02074720⟩

Share

Metrics

Record views

103