Full-waveform inversion for ground penetrating radar via ensemble Kalman inversion

Add to your list(s) Download to your calendar using vCal

• Marco Iglesias (University of Nottingham)
• Wednesday 01 February 2023, 12:30-13:15
• Seminar Room 1, Newton Institute.

If you have a question about this talk, please contact nobody.

RNTW01 - Rich and Nonlinear Tomography (RNT) in Radar, Astronomy and Geophysics

We develop a Bayesian full-waveform inversion approach for ground penetrating radar. The approach uses waveform measurements to infer the geometry of unknown metallic targets embedded in a host medium. The geometry of targets is defined via the truncation of a levelset function that we infer together with the material properties of the medium. We employ a forward modelling setting where the unknown parameters (level-set and material properties) are inputs for the open-source software gprMax which, in turn, allows us to simulate electromagnetic wave propagation in dispersive media and has built-in models of real antennas for ground penetrating radar applications. The proposed Bayesian full-waveform inversion approach is based on an Ensemble Kalman Inversion (EKI) algorithm that produces sampling approximations of the posterior of the unknown level-set and material properties. Numerical experiments with synthetic data show that, under relatively small measurement errors, EKI can accurately reconstruct the geometry of the targets within the regions of high probability produced by the algorithm. We also show numerical results with real data from an experiment where steel rebars were embedded in a concrete slab. In this case, the proposed approach was able to locate the rebars but with high uncertainty due to large modelling errors of the simulation framework.

This talk is part of the Isaac Newton Institute Seminar Series series.

This talk is included in these lists:

• All CMS events
• Featured lists
• INI info aggregator
• Isaac Newton Institute Seminar Series
• School of Physical Sciences
• Seminar Room 1, Newton Institute
• bld31

Note that ex-directory lists are not shown.