Profile Reconstruction Utilizing Forward-Backward Time-Stepping With the Integration of Automated Edge-Preserving Regularization Technique for Object Detection Applications

Yong, G. and Ping, K.A.H. and Sahrani, S. and Marhaban, M.H. and Sariphan, M.I. and Moriyama, T. and Takenaka, T. (2017) Profile Reconstruction Utilizing Forward-Backward Time-Stepping With the Integration of Automated Edge-Preserving Regularization Technique for Object Detection Applications. Progress In Electromagnetics Research M, 54. pp. 125-135. ISSN 1937-8726

Preview

PDF Profile Reconstruction Utilizing Forward-Backward Time-Stepping (abstract).pdf Download (46kB) | Preview

Abstract

A regularization is integrated with Forward-Backward Time-Stepping (FBTS) method which is formulated in time-domain utilizing Finite-Difference Time-Domain (FDTD) method to solve the nonlinear and ill-posed problem arisen in the microwave inverse scattering problem. FBTS method based on a Polak-Ribiète-Polyak conjugate gradient method is easily trapped in the local minima. Thus, we extend our work with the integration of edge-preserving regularization technique due to its ability to smooth and preserve the edges containing important information for reconstructing the dielectric profiles of the targeted object. In this paper, we propose a deterministic relaxation with Mean Square Error algorithm known as DrMSE in FBTS and integrate it with the automated edge-preserving regularization technique. Numerical simulations are carried out and prove that the reconstructed results are more accurate by calculating the edge-preserving parameter automatically.

Actions (For repository members only: login required)

View Item