Fully polarimetric radars have been widely exploited for improving target detection and classification performance. In addition, fully polarimetric Synthetic Aperture Radar (SAR) systems have been introduced to enable new applications in radar imaging. Although literature is lacking about Polarimetric ISAR systems, a growing interest for this research field is inducing many researchers to apply radar polarimetry concepts to the ISAR systems with the aim to improve image formation process, target detection and recognition.
Specifically, the polarisation state of the wave backscattered by the target depends on the target physical properties, therefore the information associated with the signal polarisation provide additional information associated with the scattering mechanisms, allowing a better understanding of the detected target. Furthermore the additional information contained in fully polarimetric data with respect to single polarisation data can be exploited for achieving better results in terms of image autofocusing.
The main idea of the work of this thesis is to demonstrate that the use of polarimetric data improves the ISAR image formation process and enhances the ATR performance.
Specifically, two innovative ATR algorithms have been proposed that make use of polarimetric ISAR (Pol-ISAR) images. In order to reduce the structure of the classifiers and, accordingly, to save computational resources, in terms of memory and computation time, both the algorithms make use of a set of features extracted from Pol-ISAR images. For this purpose an algorithm able to select the features that are representative of the target has been developed.