Research On Key Techniques For Interferometric Synthetic Aperture Radar Measurement

Interferometric synthetic aperture radar(In SAR)has become a useful technique for the generation of digital elevation models(DEMs)and also for the observation of topography deformation in nearly all weather conditions.Conventional synthetic aperture radar(SAR)measurement acquires the fine resolution of the range and azimuth directions by using the techniques of pulse compression and matched filtering,respectively.In SAR uses two complex SAR images of the same scene taken from slightly different observation positions to generate an interferogram,and with the knowledge of the interferometer geometry,phase differences between the received signals can be converted into a digital elevation model or topography change map.Image registration,phase filtering,and phase unwrapping are the key steps in the flowchart of In SAR signal processing.This dissertation focuses on subpixel image registration and phase filtering.Precise image registration presents a tradeoff between the registration accuracy and the computational complexity.Analytic subpixel offset estimation methods are proposed to solve the problem.Some technical problems in phase filtering are described and two adaptive phase filtering methods are proposed to effectively suppress the noise of In SAR phase images and preserve the detailed fringe information.The main works of the dissertation can be summarized as follows.1.Several match measures commonlly used for SAR image registration are introduced.Then we coarsely register the master and slave SAR images from X-SAR real data by the cross-correlation algorithm,and perform the fine registration by the point projection algorithm based on the principle of maximum Signal-to-Noise Ratio of spectrum,with the consistency check employed by exchanging the master and slave images and verifying the accuracy of control-point pairs.The experimental result shows that,to a certain extent,the fine registration can improve the coherence.2.Two analytic search methods based on different measurement are proposed for subpixel offset estimation.After analyzing the interpolation-typical registration methods to subpixel-level,we get the conclusion that they are all essentially discrete search methods.The registration accuracy depends highly on the size of the interpolation unit,and the registration error cannot be completely eliminated.Furthermore,performing the interpolation before the computation of matching measure leads to a large amount of computational load,particularly when high accuracy is demanded.In this dissertation,an analytic search subpixel registration method is proposed to overcome the limitations of the discrete search methods.This method establishes a novel analytical cost function by integrating computation of cross correlation(CC)and the interpolation process.Subsequently,the gradient of the cost function with respect to the offsets can be derived directly.The cost function can be effectively optimized by employing the famous quasi-Newton methods with fast convergence.Since the subpixel offsets associated with the cost function is searched in the continuous domains,this method avoids the registration error introduced by the interpolation unit used in discrete search methods,and thus improves the offset estimation accuracy.Furthermore,an efficient offset estimation method based on the empirical signal-to-noise ratio(SNR)is proposed.First,a cost function continuously varied with offsets is established by integrating the empirical signal-to-noise ratio and the interpolation operation.Second,an efficient bi-iterative algorithm is employed to solve the cost function in the continuous domains.The subpixel offsets can be exactly obtained with low computational complexity.Both the simulated and real data are tested to illustrate the good performance and computational efficiency of the two proposed methods.3.In SAR interferograms have distinct directional characteristics,which provide important information for phase filtering.A locally adaptive directional filter is proposed.The normal orientation of local phase fringes is continuously searched,and the directionally dependent filtering window is located perpendicular to the normal orientation of local phase fringes,making the pixels included in the filtering window have approximately the same elevation.The filtered phase is adaptively computed according to the minimum mean-square-error principle.The experimental resuts show that,the filter reduces the phase noise efficiently and properly avoids losing the resolution of interferograms.4.In order to effectively suppress the noise of In SAR phase images and preserve the detailed fringe information,an adaptive phase filtering method based on local slope compensation and the Anisotropic Gaussian Filter(AGF)is proposed.First,the topography-induced phase is approximately measured by local frequency estimation and removed from the original phase to eliminate the effect of the terrain topography.Second,the AGF with adaptive scale and orientation is developed to directionally filter out the noisy phase for the pixels with more homogeneous phase values.The scale of the AGF varies adaptively with the local coherence: a large-scaled AGF can better smooth the noise of low coherence areas,whereas a small-scaled AGF can better preserve the phase details of high coherence areas.Moreover,the orientation angle of the AGF is fast determined according to the maximum weighted coherent summation principle and by two-dimensional polar Fourier transform.The experimental results obtained via simulated and real data show that compared with commonly used filters,the proposed method achieves better performance in terms of residue reduction and fringe preservation.