Quantitative millimeter wave imaging of 2.5D inhomogeneous objects

A two-and-a-half-dimensional (2.5D) quantitative millimeter wave imaging algorithm, based on Newton-type optimization, is presented. The goal is to reconstruct the complex permittivity of a long inhomogeneous (lossy) scatterer, embedded in free space, from a number of time-harmonic scattered field measurements. This 2.5 D technique exploits the two-dimensional nature of the scatterer, while safeguarding the fully vectorial three-dimensional nature of the incident fields, typically Gaussian beams. The inverse problem is solved with a modified Gauss-Newton non-linear optimization technique with line search, where a multiplicative smoothing regularization is applied to the least-squares data fit. We present preliminary reconstructions from synthetic data, where 30 dB white Gaussian noise is added. This way, the effect on the reconstruction of the type of incident field (plane wave versus Gaussian beam) and of the polarization of the incident field (TE versus TM or both) is studied.