Study On Key Techniques Of Interferometric And Polarimetric Processing For High-Resolution Spaceborne SAR

The interferometric processing of SAR data can be used to measure the terrain height and monitor the movement of sea ice and currents, etc. The polarimetric processing of SAR data can be used to classify the landcover and measure the vegetation height and soil moisture. Recently, the resolution of the spaceobrone SAR systems is becoming higher and higher, and the processing of high resolution SAR data is faced with the dramatic increase in data scale, and the performance loss of conventional algorithms, etc.This dissertation is dedicated to improve the efficiency and accuracy of the interferometric and polarimetric processing of high resolution spaceborne SAR data. The major contributions are as follows.(1) The recent developments and current status of spaceborne SAR systems and the interferometric and polarimeteric processing of high resolution SAR data are summaried, both at home and abroad. The basic principles of SAR interferometry and polarimetry are presented in this dissertation. Moreover, several essential concepts in interferometric and polarimetric SAR processing are explained, including zero-doppler time system, height of ambiguity, coherence, polarization of electromagnetic wave, scattering matrix, polarimetric covariance and coherency matries.(2) Precise coregistration of high resolution complex SAR images. The studies with respect to the coregistration of high resolution SAR images are listed in the following.The conventtional coregistration methods based on the SAR geometry and the coherent/incoherent cross-correlation of SAR images are reviewed first. For the geometrical coregistration, the variation of the two-dimensional(2D) shift errors along with the terrain height is negligible, and the variation of 2D shift errors with azimuth time and slant range can be expressed in the form of polynomials. For the coherent/incoherent cross-correlation coregistration, the accuracy of the estimated shifts is heavily dependent on the data quality, and the accuracy is very high when the data quality is high. Therefore a novel method, which combines the advantages of both geometrical coregistration and cross-correlation coregistration, is proposed here. In this method, the 2D shifts of gridpixels are estimated by the cross-correlation method, and the the shift errors of geometrical coregistration are modeled and compensated by using the shifts of grid pixels in high coherence. Therefore the accuracy of the estimated 2D shifts is improved, and the effectiveness of the proposed method is evaluated by the repeat-pass interferometric data taken by Terra SAR-X.An efficient method for spaceborne SAR geolocation is devised in this dissertation, which can be applied to SAR image geometrical coregistration to improve the operating effciency. In the method, the pixels in the SAR image are divided into two types by a 2D grid, i.e., grid pixels and non-grid pixels. Then the position of the non-grid pixel is acquired by adding the position of the adjacent grid pixel to the position increment between the two pixels, where the three-dimensional(3D) position increment is computed by means of three groups of increment formulae. For geometrical SAR image coregistration, the 2D shifts of the non-grid pixel can be expressed as the sum of shifts of the adjacent grid pixel and shift increments between the two pixels, which are computed also by the three increment formulae. The experiments with simulated data and Terra SAR-X images are conducted to evaluate the effectiveness of the proposed method.(3) Noise filtering is a crucial step in In SAR processing, and its performance will affect the accuracy of digital elevation model(DEM) or measurement of movement directly. The studies concerning to improve the interferometric filtering are listed as follows.The requirements and evaluation of interferometric filtering are analysed, and a method for local fringe frequency estimation based on Fast Fourier Transform(FFT) is presented. The knowledge of local frequency enables to compensate the phase slope and therefore make the interferometric phase in the window constant. Furthermore, a coherence-weighted optimum interferometric filtering method is proposed, It is proved that the optimum weight is the reciprocal of noise variance in the meaning of least variance, and the approach to estimate the optimum weight via coherence is presented.The high-resolution interferograms over vegetation or urban areas are heterogeneous, which will violate the local stationarity assumption and make the conventional filters difficult to obtain a large number of independent and identically distributed(IID) samples. A refined nonlocal filter is proposed in this dissertation, where the similarity distancebetween the central pixel and the other pixels in the searching window is measured by the normalized probability density function of the interferogram. Then, the outliers whose normalized probability density is smaller than the given threshold are removed from the filtering process. Both the simulated and real interferograms are used to evaluate the effectiveness of the proposed method.(4) SAR interferometry includes across-track interferometry(XTI) and along-track interferometry(ATI), which can be used to measure the terrain height and sea ice movement respectively. The studies with respect to the two applications of In SAR are listed in the following.Due to the heavy computational load in locating the target by solving the In SAR geolocation equations directly or iteratively, it is important to investigate the efficient method for DEM generation in the across-track In SAR engineering. In this dissertation, the aforementioned increment formulae are extended to the DEM generation, and the experimental results show that the method can increase the efficiency of monostatic or distributed In SAR geolocation dramatically, with the loss of accuracy negalible.Monitoring of sea ice plays an important role in marine management and ship navigation, etc. For the hybrid along- and cross-track interferometric data, the interferometric phase is composed of the along- and cross-track components, and the monitoring of sea ice requires to extract the ATI phase. In this dissertation, coastlines are extracted by means of SAR geolocation and an external global DEM is used. After removing the flat earth phase, the cross-track interferometric phase of sea ice will be estimated using those of coastline pixels identified by the DEM-assisted coastline extraction method. The Tan DEM-X interferometric data of Liaotung peninsula, China, is used to demonstrate the feasibility and effectiveness of sea ice monitoring by the along-track In SAR.(5) Estimation of polarimetric covariance matrix with high accuracy. Polarimetric covariance matrix is important estimate in polarimetric SAR(Pol SAR) processing. A method of high accuracy for polarimetric covariance matrix estimation based on multi-criteria is proposed in this dissertation. In the method, the criteria of strong point preservation, polarimetric scattering preservation and back scattering preservation are considered simultaneously, and the outliers are removed according to the predeterminedthreshold in order to improve the performance. The L-band and X-band full polarimetric SAR data are used to evaluate the effectiveness of the proposed method. The experimental results show that the proposed method can suppress the speckle effect and maintain the scattering properties of strong point targets effectively. In addition, compared with the refined Lee Pol SAR filter, the proposed method can preserve the average argument of the off-diagonal terms and reduce the standard deviation of the phase noise more effectively.