Research On Several Issues Of Polarimetric Sar Speckle Reduction

Polarimetric SAR (POLSAR) has become one of the focuses in the radar system and technology field, since it can obtain the total polarimetric information of terrain. It has more and more applications, such as agriculture, cartography, disaster management, oceans, and so on. However, there is plenty of speckle in the POLSAR data, which becomes a serious impediment to the information extraction and application of the POLSAR data. Current speckle filtering algorithms have not met the demands of speckle reduction and texture protection. In view of this, the thesis takes POLSAR speckle filtering algorithm as its main topic.The main results are summarized as follows:1.This thesis defines the conception of homogeneous area. General description of the homogeneous area selection, which is raised from the perspective of pattern recognition. Studies the two current schemes of homogeneous area selection. Firstly,setting a representative algorithm of the homogeneous area selection, which is based on Prewitt edge detection oprator. The thesis summarizes the operation of the homogeneous area selection scheme based on image texture, and analyses its bugs. Secondly, basing on the homogeneous area selection scheme of Span image. it establishes the polarimetric figure space of the POLSAR amplitude data, on which defines the distance between different POLSAR pixels, analyses the polarimetric characteristics of the Span value, and disscuses the distribution model on the space of the homogeneous areas based on the Span image.2. On the basis of analysis on the similarity measure in current homogeneous area selection, in order to improve the ability of polarimetric information maintenance, The thesis presents two new polarimetric similarity measures, which serves as the theoretical basis of the thesis. Firstly, on proposing the similarity measure of the mono-polarization SAR amplitude distance, it defines the polarimetric vector distance (PVD), and determinats the space distribution of the homogeneous area based on the polarimetric vector distance and its characteristics. Secondly, the thesis defines a new conception of polarimetric vector correlation coefficient (PVC), and disscuses its characteristics.3. Making use of the two similarity measures based on polarization vector, this paper gives two homogeneous area clustering strategies. One is based on the nearest neighbor algorithm; the other one is based on the Minimax Criterion. Furthermore, this paper puts forward two refined POLSAR speckle filterring algorithms. Comparative experiments on EMISAR data show that the modified algorithms improve the suppression performance of speckle, the maintenance of texture information and polarization information significantly. 4. The maintainance of polarization characteristic playes an important role in evaluationg the polarimetric speckle filtering algorithms. Currently, the assessment of speckle polarization characteristic maintaining is usually making subjective assessment through the consistency of polarization signatures. By means of matching coefficients, which presents the similarity of two maps, the thesis puts forward the quantitative assessment of polarization characteristics retention based on polarization signatures, which completes the speckle assessment system.