Elevation Inversion Through Multi-baseline InSAR

Interferometric Synthetic Aperture Radar(In SAR) is an important technique for obtaining digital elevation model of large area quickly. Multi-baseline In SAR can greatly improve the performance of In SAR by processing data without need of phase unwrapping and ground control points. However, most of current multi-baseline In SAR methods use the approximate geometric model in calculation, which results in lower precision. And these methods with low efficiency are easily affected by system parameters error or noise and get the wrong results. These disadvantages limit the application of multi-baseline In SAR technique.Base on rigorous geometric three-dimensional reconstruction model, this dissertation studies on multi-baseline In SAR calibration, In SAR data coregistration with large deformation and multi-baseline In SAR elevation inversion. Furthermore, combining polarimetric SAR Interferometry and parallel computing with multi-baseline In SAR, this dissertation researches multi-baseline polarimetric In SAR, and implements parallel computing algorithm of multi-baseline In SAR elevation inversion.The main work and contributions of this dissertation include:(1) Constructing interferometric geometry parameter sensitivity equations under geocentric coordinate system, this dissertation proposes a multi-baseline In SAR calibration method which obtains the precise system parameters and improves the accuracy of elevation. Considering the computation of multi-baseline In SAR data must in the same coordinate system, sensitivity equations of baseline length, baseline angle, interferometric phase, initial slant range and Doppler center frequency are derived in X, Y and Z direction in geocentric coordinate system. Sensitivity equations on ground control points and tie points of multi-baseline In SAR data are combined, and system parameters of multi-baseline In SAR data are calibrated accurately, which obtain the precise system parameters and improve the accuracy of DEM.(2) This dissertation proposed an In SAR data coregistration method under large deformation condition with improved SIFT method as coarse coregistration and analytic search coregistration method as fine coregistration. For the problems of large deformation between multi-baseline In SAR data and conventional coregistration methods cannot achieves high accuracy, this dissertation presents a coregistration method. Firstly, use SIFT method extracting feature points. Secondly, match points between two images by In SAR geometry and matching criteria, and exclude mismatch points by polynomial fitting. Finally, build the correlation function containing the unknown offset and complete the fine coregistration by iterative solution pixel by pixel. Experiments show that the method has higher coregistration accuracy. The experiment shows that the method is not only suitable for the multi-baseline In SAR data but also suitable for repeat-pass airborne/space-borne In SAR data with large deformation, and has higher precision compared with cross-correlation and spectrum maximum coregistration algorithm.(3) Construct a new three-dimensional reconstruction model suitable for multi-baseline In SAR elevation inversion. Propos a multi-baseline elevation inversion method which solves absolute interferometric phase and the target coordinates, thus avoiding the phase unwrapping. Because of connecting each interferometric data by three-dimensional coordinates, the proposed method has good stability and high precision, and can withstands a certain system parameters error and accidental error, such as noise. Experiments show that, the proposed method can be used in different terrain environment, effectively reducing the accidental error while not losing detail information, and also suitable for both airborne and spaceborne data.(4) For the first time, introduce the ME-MSM and MB-ESM multi-baseline polarimetric coherence optimization method into multi-baseline In SAR elevation inversion, to improve the accuracy of DEM. Multi-baseline polarimetric coherence optimization method ensure multi group In SAR data achieve the coherence optimization at the same time, especially the ME-ESM method guaranteeing the same target in different images has the same scattering center, which is important for improving the elevation inversion quality of multi-baseline In SAR. This dissertation, through a combination of different methods, uses 15 kinds of methods on a large number of experiments, discusses the influence of various methods of elevation inversion, that MB-ESM method for multi baseline elevation inversion obtained the best effect.(5) Implement parallel computing algorithm of several multi-baseline In SAR elevation inversion methods. To deal with the problems of large amount of multi-baseline In SAR data and long computing time, the multi-baseline elevation inversion methods referred to this dissertation are transformed in parallel computation. This dissertation, to make full use of multi-core CPU computing resources, improve algorithm in Open MP parallel programming environment. Experiments show that on the workstation with 4 cores CPU and the server with 8 cores CPU, the parallel program 3 or 4 times faster than serial program.(6) Proposed technical proposal of In SAR and multi-baseline In SAR elevation inversion, and developed a set of In SAR and multi-baseline In SAR software to generate DEM integrating the above algorithm. Through a lot of experiments in Ruoergai area under different terrain conditions, the accuracy, effectiveness and robustness of this scheme and software were verified.