Study Of Polarimetric SAR Speckle Filtering And Its Effect Evaluating

Polarimetric Synthetic Aperture Radar (PolSAR) can measure the polarimeitric informationof targets, which can improve the capability of terrain detection, so it has been applied widely inmilitary and civil domain. But speckle appearing in PolSAR images is an inherent phenomenadue to the coherent interference of waves reflected from many elementary scatterers, itcomplicates the image interpretation and information extraction, and reduces the effectiveness ofimage segmentation and targets detection and distinguish. Therefore, how to reduce speckle hasbeen developed as a elementary research topic in the PolSAR image processing field,synchronously, evaluating performance plays a significant role in designing and choosingspeckle filters for different application occasion. But existing speckle filters and evaluationmethods are still incomplete. This dissertation conducts in-depth research on some key problemsof polarimetric speckle filtering technology and the main contributions are as follows:1. The radical problem of PolSAR speckle filter and its effect evaluation methods arediscussed. The historical research and development circumstance of PolSAR speckle filter andeffect evaluation methods are summarized. Put emphasis on analysing problems of existingspeckle filters which need to be solved essentially and pointing out the direction of effectevaluation methods need to be explored.2. A high-resolution PolSAR images speckle filtering based on improved SDAN algorithmis proposed. In order to deal with the problem of choosing homogeneous area just according tostatistical characteristic and ignoring polarimetric scattering mechanism in SDAN-FP specklefilters, this dissertation presents an improved SDAN algorithm for high resolution PolSARimages speckle filtering. The improved approach distinguishs the main scattering mechanism ofcurrent pixel and its adaptive neighborhood based on scattering similarity parameter. Thisprocess can make these pixels in homogeneous area have the same scattering mechanism. Thespeckle filteing experiment for simulative and real PolSAR data shows that the proposedalgorithm has the well theoretical and actual performance of preservating polarimetricinformation in heterogeneous area.3. The method for evaluating polarimetric information preservation of PolSAR specklefilteing is researched. In order to evaluate quantitively the preservation performance ofpolarimetric information for PolSAR speckle filters, an effect evaluation method of PolSARspeckle filtering based on correlation coefficients between polarization signatures is presented.Then analyse this approach ulteriorly and obtain a conclusion that the evaluating result is dividedinto co-polarized and cross-polarized correlation coefficient, which evaluating result arecommonly different. So another method based on difference degree is proposed for evaluatingpolarimetric information preservation quantitively. In evaluate process, the problems of secondestimating to Kennaugh matrices and inaccurate choosing for point targets are avoided. Themethod not only achieves quantitative evaluation for the preservation performance ofpolarimetric information, but also obtains more accurate result for speckle filtes evaluationwhich are similar evaluated by polarization signature. 4. PolSAR speckle filter effect on polarimetric target decomposition is analyzed. Becausepolarimetric target decomposition has been become a significant technique for PolSAR TerrainClassification, target detection and distinguish. Firstly, using simulative data multilookprocessing effect on three kinds of terrain and mixture media which have emblematicalscattering mechanism is analyzed. Then, several typical PolSAR speckle filter effect onFreeman-Durden polarimetric decomposition parameter are investigated by ESAR data,therefore, the purpose of effect evaluation on PolSAR image application by polarimetricdecomposition parameter is reached.