Incorporation Algorithm with RPM and DBIM in Bayesian Framework for Microwave Non-destructive Testing

Abstract

Microwave non-destructive testing (NDT) is promising for non-contact and speedy survey for air cavity or metallic rust buried into concrete media in tunnel or highway. As an imaging algorithm for the above application, the distorted born iterative method (DBIM) is one of the promising options to retrieve not only target’s location but its dielectric property, which is useful for material characterization. However, in an actual NDT scenario, scattered data from limited direction are available, which makes the problem more ill-posed. In this paper, the incorporation algorithm with radar approach as the range points migration (RPM) method, which offers a prior estimate of the region of interest (ROI), substantially reduces the number of unknowns in the DBIM. Furthermore, the variational Bayesian expectation maximization (VBEM) algorithm is introduced in the above incorporation. The finite-difference time-domain (FDTD) based numerical simulations demonstrate that the proposed method achieves faster convergence and more accurate results compared with that without the prior ROI estimation.