Improved Height Estimation Algorithms for Synthetic Aperture Sonar Interferometry

Abstract

Synthetic aperture sonar (SAS) interferometry is a powerful tool for generating high resolution bathymetric maps. The basic principle of interferometry is to relate the time-delay between echoes received by two displaced sensors to a height. Cross-correlation and phase-differencing are techniques used in SAS interferometry for estimating time-delay, which both require the preceding processing steps to be phase preserving. Ground plane imaging is used to facilitate interferometry by reducing geometrical distortions and registering the sonar images. This introduces a frequency shift to the images, which again reduces the quality of the height estimate. We propose a deterministic and a statistical approach to compensate for this shift. Both the deterministic and the statistic correction perform well on both simulated test cases and on real data collected by the SENSOTEK interferometric SAS during HUGIN trials. We also propose the cross ambiguity function (CAF) as an alternative to cross-correlation for estimating time-delay. The CAF method is better suited for handling areas limited by geometric distortions (or baseline decorrelation). CAF shows better performance than cross-correlation, both on ground plane images from areas with incorrect initial height, and on images of complex objects. This is validated both on simulated test cases and on real data from SENSOTEK.