Free-space determination of permittivity, size and orientation of rectangular shaped objects using multivariate analysis

This paper presents a free-space method for the determination of the permittivity of non-dispersive low-loss rectangular shaped dielectric objects. The dielectric object under investigation is illuminated by a sub-nanosecond pulse radiated from a linear polarized transmitting antenna. A scattered pulse is received by a dual polarized line array. The object can be moved orthogonally to the line array forming a virtual two dimensional antenna array. Several objects of different shape, size, orientation and permittivity are measured. It is supposed that the time domain data contains information about the shape and the permittivity. The data is then subjected to multivariate analysis to separate those effects due to the geometry of the object from those due to its dielectric properties. Results for the determination of the permittivity independent of shape, size and orientation using principal component regression and artificial neural networks are presented. It is shown that the determination of the size and the orientation is possible to a certain accuracy as well.