UBC Theses and Dissertations
Detection of man made targets using radar polarimetry Wasniewski, Flavio
Since the late 1970's, various synthetic aperture radar (SAR) satellite missions have provided a valuable source of information about the earth's surface. Providing their own illumination, these sensors generate imagery of our planet 24 hours a day, regardless of cloud cover. Until recently, except for the short term space shuttle SIR-C system, these SAR missions carried single-polarized sensors, meaning that the information extracted from the imagery, and the resulting quality of interpretation, was limited. In 2006 the first of a series of planned satellites carrying fully polarimetric SAR sensors was launched. In polarimetric SAR systems, images can be acquired with horizontal (H) and vertical (V) polarizations of the electric field on both transmission and reception. The resulting multipolarized imagery makes available a higher level of information content for a number of applications, including man made target detection. The potential of polarimetric SAR for man made target detection is investigated in this thesis. In the man made target detection field, the purpose is to differentiate the signature of the targets from those of adjacent natural areas. The most operationally successful application in this field is ship detection, as the sea surface usually forms a quite homogeneous and easily distinguishable clutter. Vegetated surfaces, however, usually constitute a more challenging clutter, as the backscatter levels can be very high and the signatures diverse, often being confused with those of man made targets. A few algorithms have been developed with the purpose of separating the targets from clutter. A methodology that employs the following three algorithms was recently tested in order to discriminate crashed aircraft from the surrounding terrain: Polarimetric Whitening Filter (PWF), Even Bounce Analysis and Cameron Decomposition. In these tests, successful results were achieved when the terrain was composed of a homogeneous field of grass and the targets were crashed airplanes. In this work the same methodology is tested with different man made targets and with different clutters. Also, methodologies involving the use of two other algorithms are tested in order to reduce the false alarm rates. These algorithms are Coherence Test of the Symmetric Scattering Characterization Method (SSCM) and Freeman-Durden Decomposition. The methodologies are applied to three different data sets acquired over different Canadian locations - Ottawa (ON), Gagetown (NB) and Vancouver (BC) - with the CV-580 polarimetric SAR system and the false alarm rates are assessed. Results show that the Coherence Test can lower false alarm rates on high vegetation clutter when applied in combination with other algorithms, while Freeman-Durden Decomposition does not perform effectively in the same experiments.
Item Citations and Data