POLSAR Image Simulation, Detection And Classification

In the military field, Polarimetric Synthetic Aperture Radar (POLSAR) plays an important role in the application of surveillance and reconnaissance. Three critical problems, including the simulation of POLSAR image, the target detection and classification in POLSAR, are investigated.In section one, the simulation of POLSAR image is studied and the whole flow chart is presented. Based on the different statistical properties between targets and clutter, the targets' and clutter's POLSAR image are simulated respectively. Firstly, the Gaussian statistical model for taget/clutter, Gamma product model for clutter and the determinate model for target are presented respectively for the low-resolution and high-resolution cases. Secondly, a simulation method of the targets' POLSAR image, which is based on coherent video echoes, is obtained. Thirdly, the Gaussian and Gamma product clutter POLSAR image are quickly obtained through decomposing the clutter's covariance matrix. Finnally, some simulated results are presented.Especially, from the view points of antanna pattern, channel coupling and clutter simulation, the differences between the simulation of POLSAR image and that of SAR are clarified. The POLSAR image simulation is also been considered under the condition of Electric Warfare(EW). The simulation can be applied for POLSAR system design and the coherent POLSAR algorithm validation as well.In section two, the polarimetric detection algorithms are studied. Firstly, the conventional polarimetric detector such as Optimal Polarimetric Detector(OPD), Polarimetric Whiten Filter(PWF), Span detector are introduced. And the comparison of their detection performances is presented, too. Then a novel polarimetric detection algorithm based on Stokes subvector is presented. The experiment's result shows that the new algorithm's performance is improved.In section three, the problem of classification of POLSAR image is studied. Firstly, a summary to polarization Target Decomposition (TD) and the target classification algorithms based on TD is pesented. Thereby, a new characteristics selection algorithm is presented based on Partially Coherent Target Decomposition (PTCD). The rule and schedule of optimal characteristics selection is analyzed in detail. Finnally, the experiments results using the real POLSAR data of NASA/JPL AirSAR show that the classification accurate rate reaches 95.7%.