A Geometrical Autofocus Method for UAV-Based SAR

Uncrewed aerial vehicle (UAV)-borne synthetic aperture radar (SAR) radar technology is gaining increasing attention in the remote sensing field due to its low cost, high flexibility, rapid response, and short revisit time. High-quality SAR imaging requires precise platform localization; however, low-cost, lightweight UAV platforms often struggle to achieve the subwavelength navigation accuracy required, particularly at high carrier frequencies. To address this limitation, autofocusing (AF) algorithms are commonly employed to compensate for unknown trajectory deviations. In this work, we present a novel geometrical evolution of the well-established phase gradient autofocus (PGAs) algorithm, specifically designed to handle severely nonlinear trajectories in highly space-variant scenarios, as typically encountered in UAV-based acquisitions. The key innovation of the proposed geometrical autofocus (GAs) method lies in its rigorous geometrical approach to estimate the true trajectory traveled by the platform, from sufficiently distributed point-like targets in the scene, fully accounting for the 3-D nature of the acquisition. Additionally, GA is applicable to all SAR imaging modes. Based on its geometrical formulation, GA can also be employed for the simultaneous deblurring and alignment of multiple SAR images acquired from di fferent trajectories, serving as an initial co-registration step for InSAR applications. This article further demonstrates that assuming a spatially invariant phase error is not valid in UAV applications due to the low flight altitude. The proposed autofocus method is compared to the traditional PGA algorithm using both simulated and real UAV SAR data. Two real datasets were processed, the latter without calibration corner reflectors, to confirm that GA can be applied effectively even in their absence, as long as some point-like targets are present.