Backscatter Features for Estimating Synthetic Aperture Sonar Bathymetry

Synthetic aperture sonar (SAS) bathymetric mapping has matured in recent years and is adopted by industry, government and academia for mapping and exploration of large areas in more detail than previously feasible. As with other types of bathymetric mapping, averaging is often a critical processing step in order to gain sufficient signal quality to ensure the reliability of the bathymetric product, i.e. the depth map. SAS stands out from many of the other mapping technologies by the availability of very high resolution, centimeter-scale backscatter images. In this paper, we investigate the use of backscatter features to enhance the resolution and accuracy of the bathymetric product. The aim is to allow averaging over larger spatial extent inside of the interferometric processing, while preserving sharp edges and object boundaries, as well as improving the similarity of the statistical properties for the averaged cells. We suggest a method for achieving this by using image segmentation on backscatter data, and demonstrate the method on simulated and real data sets. We achieved better edge preservation on protruding objects, more accurate bathymetry estimates, higher object signal to noise ratio and sharper SAS images after geometric corrections due to the improved bathymetry estimate.