A Complete Three-Dimensional Autofocus Pipeline for Multi-Pass Backprojection SAR with Generalized Phase Gradient Autofocus, Registration, and Elevation Phase Calibration

Abstract

Backprojection synthetic aperture radar imagery requires wavelength-accurate platform positioning, as small trajectory errors introduce phase errors that produce azimuth smearing and degrade image quality. Existing autofocus algorithms such as Phase Gradient Autofocus require a Fourier transform relationship between the range-compressed data and the image domain, preventing their direct application to time-domain backprojection imagery, and no complete autofocus pipeline exists for multi-pass three-dimensional backprojection SAR. This thesis presents a complete three-dimensional autofocus pipeline built on three sequential steps: the derivation and real-data validation of the Generalized Phase Gradient Autofocus algorithm on airborne backprojection SAR data, a phase correlation registration method for correcting inter-pass geometric displacement errors that lie in the null space of the phase gradient estimator, and an elevation-domain extension of GPGA that corrects inter-pass phase calibration errors under a separable error assumption — together enabling a complete three-dimensional backprojection autofocus pipeline combining registration and phase gradient autofocus in both the azimuth and elevation dimensions.