A Study On The Techniques Of Fully Polarimetric Sar Data And Sar Compositive Processing System Implementation

Fully polarimetric SAR radar is the primary means of radar classification and recognition because it can work all-weather, all-time with the penetration ability and other characteristics. Both in the military and civilian field, fully polarimetric SAR radar has extensive and irreplaceable application market. Compare with single polarimetric SAR radar, fully polarimetric SAR radar have access to a variety of polarization combinations of the echo, and thus be more informative. Because of the significant increase in the amount of available information, polarimetric SAR can describe the features of the physical scattering characteristics, so that it can achieve more accurate classification and recognition results. Applications of classification and recognition based on polarimetric SAR have become a current research focus. Based on the characteristics of polarization SAR data, this paper research on polarization SAR image speckle reduction and polarization SAR unsupervised classification algorithm and construct the polarization SAR image algorithms library with a software that use the algorithms library. Main content in this paper is:1. Introduce the mathematical description of the polarization vector, as the basics of this paper. Analysis the SAR image speckle formation mechanism, mathematical description and statistical properties detailedly. Then summarize the classical polarization SAR speckle reduction algorithms and non-supervised classification algorithms based on polarimetric decomposition. The contents of this chapter lead to the research work at following chapter as a foundation.2. Because parameter estimation step of the speckle reduction algorithms has a decisive significance. Firstly, Summary existing parameter estimation algorithms and analyse the advantage and weaknesses of these algorithms. Then for the performance evaluation indicators of speckle reduction effect and detail retention capacity, analyze the different characteristics of textured regions and non-textured regions. Estimate parameters in two ways for different regions, then a speckle reduction algorithm based on improvement of paramteter estimation algorithm is proposed. The effect of this algorithm has both high speckle reduction and good preservable capability of image detail and polarization characteristics. Finally, we compare this algorithm with other algorithms with some the performance evaluation indicators, this experiment prove the effect of this algorithms.3. Analyze the classical unsupervised polarization SAR classification algorithm. aim to the problem of using polarization characteristics incomplete and ineffective existing in current classification algorithm, firstly, select one algorithm from each three types of decomposition algorithm, building a characteristic matrix with the polarization characteristics calculated from these polarization decomposition algorithms. Then proposed the integrated using of characteristics polarization SAR Unsupervised Classification Algorithm based on the characteristic matrix and theory of comparability of matrix. The algorithm use the characteristics calculated from different type of polarization decomposition algorithms fully, it optimize class borders through iterative clustering and improve the classification accuracy. Finally, its effectiveness is proved though compare with other Classification Algorithm.4. Based on the analysis the defect of current polarization SAR image processing software systems, construct a comprehensive and expandable multi-polarization SAR image processing algorithms library and a single-polarization SAR image processing algorithms library. Algorithm library contains SAR images filtering, classification, registration, fusion, super-resolution and performance indicators functions. On this basis, a comprehensive polarization SAR image processing software systems based on MVP platform is designed and implemented with VC6.0. The system can easily SAR image processing algorithms library we constructed and implement functions of these algorithms. A long-time using proved the practicality and effectiveness of this system.