Study On Key Techniques For Interferometric Synthetic Aperture Radar Measurement

Interferometric Synthetic Aperture Radar (InSAR) measurement is a technique that reconstructs the terrain elevation or topography change information according to the Synthetic Aperture Radar (SAR) data. Conventional SAR measurement acquires the fine resolution of the range and azimuth directions by using the techniques of pulse compression and matched filtering, respectively. By observing over the same scene at different times or along different tracks, one can obtain the complex images of the same scene on the ground. After the registration of the images, one can get the interferometric phase by subtracting the corresponding phase. There are fixed geometric relation between the object on the ground and spatial geometric framework of the interferometer. On the basis of this relations and some parameters such as platform altitude, radar wavelengh, incidence angle, baseline, and so on, one can accurately get the height and change information of the objects on the observed scene.This dissertation provides the process of signal processing for spaceborne interferometric SAR, with emphasis on the key techniques of signal processing, such as registration, phase filtering, and phase unwrapping. The main works of the dissertation can be summarized as follows.1. The principle of SAR imaging and two classic imaging methods are described. Based on it, it is also indicated that phase errors of range and azimuthal directions in SAR image pair is included in a linear phase term of impulse response function for SAR system. According to this conclusion, a new registration method based on data domain is proposed. Based on the signal model of the point target, this method gets the relative offsets by multi-look techniques and numerical operations. After compensating the offsets, one can complete the registration. This method avoids the error inducing by taking amplitude/phase-based registration methods. It does not need to the operations of amplitude correlation and interpolation. Therefore its operation is simple. Compared with conventional image registration methods, the proposed method has better performance.2. The existing registration methods based on the SAR image all employ area-based method. An image registration method based on the feature information is proposed. This method implements the pixel-level registration by using the mutual features on the image pair. In feature detection, a improved templates method is used to detect the feature. In feature matching, the process is implemented by calculating the relative distances of the feature points. This method reduces the computational load effectively in the case of assuring pixel-level registration accuracy.3. On the basis of the above method, a registration method of curve fitting is proposed. This method uses B-spline function to fit the feature curve of the image pair after edge detection. The curve matching is implemented by using the principle of weight moments and correlation matching. After finding the centroids of the curves and the whole offsets of the image pair by lease-square method, one can implement the pixel-level registration of image pair. This method has low computational load and excellent robustness.4. After analyzing the existing subpixel level registration methods, we get the conclusion that they are all essentially discrete search methods. They acquire subpixel level registration accuracy by searching the extremums of the cost function in the interpolated discrete domains. Furthermore, the improvement of registration accuracy is limit and the registration errors cannot be completely eliminated. To overcome the limitations of the discrete search methods, this dissertation proposes a analytic search subpixel registration method for interferometric SAR image. This method searches the subpixel offsets in the continuous domains, which avoids the registration error introduced by the interpolation unit using discrete search methods. The performance of the proposed method has been demonstrated by registering the simulated and real InSAR image pairs. The results of comparison with classic methods are displayed to illuminate the high accuracy of the proposed method.5. We discuss others key techniques for InSAR measurement, such as flat earth removal, phase filtering, phase unwrapping, and Digital Elevation Model (DEM) reconstruction, with emphysis on the phase filtering and phase unwrapping. In phase filtering, a locally adaptive directional filtering is implemented. This method owns a better performance in fringe preservation and residue reduction. In phase unwrapping, a effective synthetical phase unwrapping method is presented. This method, based on the branch-cut method and finite element method, combines the advantages of the two methods, i.e., in a high-quality region, unwrapped phases retrieved by the branch-cut method are more accurate than those by the lease-square method; in a low-quality region, the robustness of the least-square method is better than that of the branch-cut method. This method assures the phase unwrapping from high-quality region to low-quality region, and resumes the whole and smooth phase map. This method not only assures the accuracy of phase unwrapping in high-quality region, but also improves the accuracy of the phase unwrapping in low-quality region at a certain extent since the boundary value of estimating is reliable from high-quality region.