TECTECHNICAL DETAILS: The Complex-Valued Neural Networks (CVNNs) are computational intelligence tools that process complex-valued information by using complex parameters and variables. They are particularly useful when complex vectors of electromagnetic waves must be processed, as is the case of multi-polarization radar images. The thesis work will consist of a study of the networks, a simple implementation in IDL environment and of an engineering evaluation of the performance of the developed tool applied to radar measurements on forests.

FRAMEWORK: Several satellite missions, one of which is the Italian COSMO, are being launched to observe the Earth with new-generation radars. The radar images find applications in many fields, but, unlike the optical case, special processing is required by the complex (amplitude and phase) nature of the observed quantities. A particularly important application is the radar interferometry, from which the 3-D structure of the earth surface and of the objects can be determined. The figure illustrates the determination of the 3-D structure of a forest obtained from measurements by the German Aerospace Center (DLR) multi-polarization radar.
PREFERRED SKILLS: Average-good computer skills, good understanding of electromagnetic fields and knowledge of English are desirable.
OFFERED BENEFITS: The selected student will learn the fundamentals of IDL programming and will acquire a good overview of Neural Networks.