Research On Synthetic Aperture Radar (SAR) Image Matching For Navigation

Image matching for navigation is an important technology in automatic aircraftnavigation with high accuracy. Although many Aircrafts navigate automaticly based onvisible-band sensors, they are always influenced by weather and lighting condition, thuswork with a low efficiency. Synthetic Aperture Radar(SAR) is an important development inmicrowave remote sensing in modern remote sensing technology,as an active imagingsensor,SAR can not only provide sufficiently high resolution images as visible-band ones,but also work all-weather, day-or-night without being influenced by weather. Because ofthese advantages, SAR is a very good sensor for autonomous navigation and guidance. This dissertation addresses the technology of SAR image matching for automaticaircraft navigation and conducts deep investigation in the following key technologies: (a)SAR image speckle denoising; (b) SAR image matching technique with high performance;(c) Matching SAR image to satellite visible-band one based on common feature; (d) 3Dterrain matching technology based on synthetic aperture radar interferometry (INSAR).Thefollowing are the details. (1) The principle of SAR is briefly introduced in the dissertation. INSAR technologywhich is used to produce terrain data is also introduced, together with the property of SARimage. (2) Addressing SAR image speckle denoising, this dissertation proposes a new methodbased on statistical model of wavelet coefficients combined with modification to themaccording to significant coefficient rule. In the method, wavelet coefficients of image arefirstly modeled as mixture density of two Gaussian (MG) distributions with zero mean. Inorder to incorporate the spatial dependencies into the denoising procedure,Hidden MarkovTree model (HMT) is explored and Expectation Maximization (EM) algorithm is proposedto estimate model parameters.Bayes Minimum mean square error (Bayes MMSE)methodis used to estimate the wavelet coefficients free of noise. The wavelet coefficients areupdated according to a rule whether the coefficient is a significant one or not.2D inverseDWT is performed on the updated coefficients to get denoised SAR image. ExperimentalResults using synthetic and real SAR image demonstrate that the method can not onlyreduce the speckle but also preserve edges and radiometric scatter points. (3) A new method for high resolution SAR image matching is proposed using Pyramidmulti-resolution algorithm in wavelet domain combined with information fusiontechnology in the dissertation. In this method, pyramid multi-resolution algorithm is IIIemployed to make matching algorithm faster. Wavelet transform is used to constructmulti-resolution pyramid in that different subband corresponds to different characteristic ofimage (approximate components and edge of different directions). In order to improve thematch probability and location accuracy, the matching results of all subbands are fusedusing Dempster-Shafer rule to get the finer results. Experiments under noise of differentlevel,geometry transformation and different wavelet base function are conducted to test theperformance of the matching algorithm, and some useful conclusion are drawn.Thereal-time algorithm based on the method has been implemented in micro-processor. (4) With the development of aircraft navigation, more and more attentions are paid tomatching images of different sensors to get higher location accuracy. The technology ofmatching SAR image to satellite visible-band one based on common features is addressedand a new method is proposed in this dissertation. In this method,the edge comes as thecommon feature, and an edge extraction algorithm is proposed according to the property ofSAR image. A new similarity measure based on Hausdorff Distance Matrix FrobeniusNorm is proposed. This measure is more reliable and can achieve higher location accuracycompared with other measures based on Hausdorff distance u...