On the relationship between SAS image resolution and target-detection performance

Abstract

The relationship between synthetic aperture sonar (SAS) image resolution and target-detection performance is quantified. It is first demonstrated how a lower-resolution SAS system can be simulated in a principled manner by band-limiting the image wavenumber spectrum of a full-resolution SAS image. This insight enables a direct comparison of target-detection performance in imagery of different resolutions while all other variables (such as data-collection conditions) remain fixed. The comparison is achieved by degrading a given SAS image to a series of different along-track resolutions and range resolutions, and then applying the same detection algorithm to the resulting set of images. This is repeated on a large scale for a set of real SAS data collected at sea at eight different geographical locations in various environmental conditions; the data considered comprise over 1500 images collectively spanning over 8 square-kilometres of seabed and containing over 2000 mine-like targets. The results quantify the dependence of target-detection performance and false alarm rate on along-track resolution and range resolution. This information can be used to determine the minimum processing requirements needed on board an autonomous underwater vehicle for effective through-the-sensor survey adaptation, and also to inform SAS system design and procurement decisions.