The Phase Gradient Autofocus Algorithm for Strip-Map SAR Data

Abstract

The Phase Gradient Autofocus (PGA) algorithm was developed for spotlight mode SAR for phase error estimation. Different approaches of this algorithm have been investigated in the frame of this thesis. First different kernels of the PGA estimation are presented: the original kernel of the PGA algorithm [1-2] and the Maximum Likelihood estimation [3]. The aim of this work is to implement PGA for strip-map SAR images. Different from the spotlight case the phase history of different scatterers in the strip-map image is not the same, making the implementation of PGA more complicated and with the same drawbacks with respect to the linear phase estimation as we will show in our implementation. Since strip-map SAR systems, typically, are operated at low altitudes (e.g. the E-SAR system), we have extended the PGA algorithm to the range dependent case. Additionally, the Phase Weighted Estimation (PWE) is incorporated for more robust estimations. These algorithms are tested with simulated point targets in white Gaussian noise, and by adding synthetic phase errors in real strip-map data (E-SAR data of Oberpfaffenhofen) . The data are corrupted by several phase errors. These phase errors have different characteristics (linear, quadratic, etc…) and different amplitudes in radians. The adapted implementation gives good results on areas with corners reflectors, but it must be improved and extended by iterative steps to allow its application to a complete scene. This work is divided as follows: Section 2 of this report discusses the basics of the Phase Gradient Autofocus for Spotlight SAR imaging. Section 3 describes our implementation of PGA for strip-map SAR data. The phase estimation results are presented in section 4 and the future research is discussed in the conclusion section.