Estimating layover in broadband synthetic aperture sonar bathymetry

In this paper we investigate broadband interferometric height estimation using a small vertical hydrophone array and multiple frequency bands as a means of reducing image layover artefacts. We demonstrate the effect of layover on bathymetric images using simulated imagery to avoid the ground-truth problem. We then discuss maximum likelihood (ML) methods for estimating multiple targets within the same resolution cell and extend these to multiple frequency bands. Results are shown for two frequency bands (20-40 kHz and 90-110 kHz) and a vertical hydrophone array of three elements as used by KiwiSAS. Each of these bands is subdivided into a number of sub-bands to reduce the footprint shift effect, or misregistration between the different signals, due to the unknown seafloor topography. Multiple spatial looks are also employed to reduce the variance of the phase estimates. A relaxation algorithm is then applied to estimate multiple targets within a resolution cell. The algorithm was found to give reasonable estimates except when two (or more) closely spaced scatterers of comparable strength were in the same resolution cell.