Development of Advanced Co-registration Methods for Sentinel-1 TOPS Mode

Abstract

Spaceborne synthetic aperture radar systems have been very important for Earth monitoring since the launch of the SEASAT. Wide-swath mode SAR is one of the most important aims of the SAR systems that provide earth surveillance with a short revisit time at the cost of azimuth resolution reduction. Terrain Observation by Progressive Scans (TOPSAR) is a recently proposed wide-swath mode SAR that is expected to replace the ScanSAR mode. TOPSAR's advanced acquisition technique overcomes the drawbacks of the ScanSAR mode, such as scalloping and azimuth-varying signal-to-ambiguity ratio. However, the squinted acquisition geometry of ScanSAR and TOPSAR introduces a phase ramp in both azimuth and range directions. The phase ramp in range direction is negligible whereas the phase ramp in azimuth direction remains due to the azimuth varying Doppler centroid frequency. For the interferometric applications of TOPSAR mode data, the phase ramps of the SLC pair require a co-registration accuracy in the milli-pixel order. Enhanced Spectral Diversity (ESD) is a co-registration method that exploits the overlapping area between two consecutive bursts to estimate the mis-registration with the required sensitivity. However, a prior fine co-registration step is a prerequisite for the ESD. In this thesis, performance of the ESD is verified and a robust mis-registration estimator based on ESD method is proposed. According to the TOPSAR scene geometry, the overlapping area types of burstwise, beamwise-forward and beamwise-backward are introduced. Simulated data as well as Radarsat-2 data take over Salar de Uyuni are used to assess the performances of overlapping area types. Results show that the azimuth shift estimation accuracies of all overlapping area types are consistent with the theoretical expectations. The estimation sensitivity of the burstwise overlapping area is better than that of the beamwise overlapping areas. The proposed estimator combines all overlapping areas in the scene to exploit different spectral separations and increase the number of the samples in order to relax the fine coregistration requirement of ESD and improve the estimation sensitivity. Sentinel-1 data takes over Genoa, Munich and Istanbul are used to investigate the estimator's capabilities. Results prove that the estimator provides sensitive mis-registration estimation for TOPSAR IW data without a fine co-registration step in case of a mis-registration less than half a pixel.