Study Of Polarimetric SAR Remote Sensing Information Processing Based On Polarimetric Decomposition

The polarimetric SAR technique,which takes different polarizations for transmission and reception,has received a considerable amount of attention for its capabilities concerning information acquisition.This thesis focuses on polarimetric SAR information processing and image interpretation.Major works are arranged as follows:1.The generalized Stokes vector in three-dimensional case is derived by representing the electric field density matrix with Gell-mann basis belonging to the SU(3)group,inspired from the deduction of Stokes vector in two-dimensional case by representing the density matrix with Pauli basis belonging to the SU(2)group.Furthermore,the covariance matrix in bistatic case is represented with the generalized Pauli bases belonging to the SU(4)group,and the corresponding coefficients are proved as elements of Mueller matrix,which verifies that Mueller matrix can be regarded as the generalized Stokes vector in the SU(4)group.2.The degree of polarization(DoP)of n-dimentional density matrixes is obtained from the relationship between eigenvalues of density matrixes and prolarization degree,so that the Degree of Polarization can be calculated directly from covariance matrix and coherency matrix.The further analyis shows that DoP can be understood as the distance between the normarlized density matrix and identity matrix,i.e.,the Degree of Polarization of a density matrix is proportional to its differenc with the identity matrix.3.The polar decomposition of Sinclair matrix is invesgated in coherent situation,and the significances and functions of unitary matrix and Hermite matrix operators are discussed.For the incoherent situation which is more generalized,the improved polar and symetrical decompositions for Mueller matrix are proposed.Parameters derived from these two decompositions are applied to terrain classification,ship detection and oil spill surviliance.4.The three-component model under three typical compact polarimetric modes,CTLR,DCP and ?/4,are derived through projecting the three-component model under fully compact mode to these three modes.A modified Freeman-Durden decomposition method is proposed for compact modes.To solve the underdetermined equations for compact polarimetric decomposition,the contribution of volume scattering is taken as a free parameter,thus giving a complete set of solutions in an explicit format.The estimation of volume scattering contribution is then derived,within the constraint that the weighting factor of each scattering component should be nonnegative.The proposed decomposition is rewritten in the form of Stokes vector,and the physical implication is discussed.5.The proposed compact polarimetric SAR decomposition is extended to the general case with elliptical transmition and linear reception(GTLR mode).The optimal compact polarimetric mode is investigated from the decomposition quality under different elliptical parameters.The proposed technologies are all successfully applied to the polarimetric SAR field data processing.