Generation and use of optimised databases in microwave characterisation
Résumé
A method for optimising the sampling points in a database used to characterise a coaxial probe for dielectric measurement is described. The optimised databases are related to the forward operator that maps the model parameters (the material dielectric properties) to the observation (measured data). The optimised sampling method is shown by simulation to improve the accuracy of dielectric measurements, in the presence of measurement noise, when a neural network is trained using the database compared with a regularly sampled database.