Wideband Interferometry in Synthetic Aperture Sonar

Abstract

Synthetic aperture sonar (SAS) interferometry can provide very high resolution images and topographic maps of the seafloor over large swaths. Processing of interferometric data to retrieve reliable depth estimates of the seafloor is, however, nontrivial. Traditional narrowband interferometry processing relies on advanced phase unwrapping techniques, constraints, and prior knowledge to resolve phase ambiguities. These methods all have dependencies throughout the dataset, e.g., a small error in the assumption may cause a global error in the final phase estimate. For wideband systems, there are alternative techniques to estimate the absolute (or ambiguity-free) phase difference directly. We consider four different wideband interferometry techniques for direct phase difference estimation: complex cross correlation, split-spectrum algorithm, and the multichromatic approach. In addition, we develop a weighted split-spectrum algorithm, where the weights minimize the variance in the absolute phase estimate. We benchmark these techniques on simulated data and demonstrate the techniques on real data from the HISAS wideband interferometric SAS on a HUGIN autonomous underwater vehicle. We have found the following: the cross correlation technique always outperforms the other techniques in misregistrated areas caused by severe topographic changes. The split spectrum techniques are substantially faster than the cross correlation technique. The multichromatic approach has similar performance as the multiband split spectrum technique for the same choice of bands. We demonstrate that all the wideband interferometry techniques outperform a standard Goldstein's Branch cut phase unwrapping algorithm on real data from a complicated scene containing an elevated target and severe layover.