The Application Of Polarimetric Information For SAR Targets Detection And Classification

Owing to its capability for obtaining the complex scattering matrix of targets,polarimetric synthetic aperture radar(PolSAR)can provide more information about targets,leading to an accurate targets locating and a possible category classification.Focusing on the ships at sea,this thesis mainly study the application of polarimetric information for ships detection and classification.The main contributions can be summarized as follows:1)To discriminate the cross-polarized scattering power produced by oriented scatterers from that by volume scatterers,and to take fully account of ships' reflection asymmetry property,a cross-polarized scattering model is proposed;then combining it with the entropy one volume scattering model,the surface and double-bounce scattering models,a four-component model-based decomposition method is developed for accurately analyzing ships' scattering mechanisms;2)A new PolSAR ship detection metric fused by polarimetric similarity parameters and the third eigenvalue of the coherency matrix is proposed.It can not only enhance the ship-sea contrast significantly,but also have the capability to suppress the sidelobe of ships and other disturbances.Based on the metric proposed,a clustering-based detection method is designed further and applied to ship detection successfully;3)Three polarimetric features,i.e.,the mean scattering mechanisms complexity(MSMC),the average total power(TP)and the average total power encoding(TPenco)features are proposed to characterize ships'scattering mechanisms and scattering power.Based on the SVM classifier,experiments by using the proposed polarimetric features and several non-polarimetric features including geometric features,texture features and invariant features are conducted.Experimental results show that the scattering mechanisms and the scattering power features are both effective for ship classification,which implies that it's worth further studying on applying the polarimetric features to ship classification.