|
||||||||||
Ana Carpio, Applied Mathematics - Optimization, inverse problems and imaging |
||||||||||
PDE constrained optimization methods
in conjunction with topological sensitivity based
techniques provide a powerful tool to reconstruct the
geometry and properties of objects buried in a
medium by scattering of electromagnetic,
acoustic or thermal waves. The idea is to exploit
the topological sensitivity of cost functionals
defined in terms of the data measured at a set of
receptors to produce first guesses of the objects.
The procedure is robust to noise. When enough
incident directions or frequencies are employed, a
topological derivative based iteration may provide a
reasonable description of the objects
geometry. The method works with both harmonic
or general time dependent incident radiations by
either selecting different incident directions or
recording information at different times. We have
applied this strategy to acoustic sounding of
bodies, electrical impedance tomography and
photothermal imaging, combining it with gradient
techniques to identify material parameters. We are
able to generate good first approximations of the
number, size and location of the scatterers, and
improve the description of their shape and material
parameters in a few steps. Making use of iteratively
regularized Gauss-Newton techniques, we can invert
holographic data in noninvasive light imaging of 3D
biological samples. Resorting to bayesian
approaches, and either linearized posteriors about
an optimal point or Markov Chain Monte Carlo
sampling, we are able to quantify uncertainty in the
inversion process for increaing noise magnitudes.
|