Statistical Study of SAR Transionospheric Autofocus Procedure

Synthetic aperture radar (SAR) has been the imaging resource of choice for decades due to its all-weather capabilities and independence from external illumination. Orbital satellites that emit long-wavelength signals have the capacity to “see” through vegetation, snow, and ice. However, these systems are affected by the phase perturbations due to the ionospheric turbulence that phase shift the propagating radar signals, causing image distortions and blur. This study is a continuation of our previous work where we have developed an autofocus algorithm to improve the image quality deteriorated by the ionosphere. Here we conduct a statistical analysis of the efficiency of optimization-based SAR autofocus algorithm developed in our earlier publications. In particular, we test the limits and sensitivity of our autofocus algorithm that includes a gradient-based optimizer applied to a contrast-enhancing cost function. The role of perturbation factors, such as the intensity of the ionospheric turbulence, the average clutter reflectivity, and the average level of background noise, is investigated. Using the advantage of computer simulation, we are able to compare the initial corrupted image, as well as the autofocus outcomes, to the “true” image obtained with the perfect compensation of ionospheric distortions. The autofocus effectiveness is evaluated using three image quality metrics: normalized cross correlation (NCC), integrated side lobe ratio (ISLR), and the desynchronization of peak locations (PD). We find that our autofocus algorithm is capable of improving image quality and achieving a high level of similarity to the true image in most situations, with the high clutter level identified as the most challenging part of the setup.