The SAR map-drift algorithm extended for a multi-hydrophone SAS

The problems caused by random sway errors in the tow-path of a single hydrophone, synthetic aperture sonar (SAS) are well-known. If these horizontal displacement errors are left uncorrected and the tow path is assumed to be straight, they have a devastating effect on the quality of the reconstructed image. Although on-board navigation instruments can help measure the gross departures from the straight path, sub-wavelength residual sway errors still corrupt the imaging process. An early sway estimation technique in SAR (called the map-drift algorithm) was to divide the full aperture up into two equal, but non-overlapping sub-apertures. By cross-correlating the resultant two low-resolution images, an estimate of the equivalent phase error (that is the sway error divided by the wavelength and multiplied by /spl pi/) could be obtained for the second order coefficient of the polynomial describing the sway between the images. The two-image map-drift algorithm could be extended to smaller and smaller sub-apertures with the phase error estimated to higher and higher orders of the polynomial. The limit of the method is when the sub-aperture images are of such a low resolution that it is not possible to detect the differential sway between adjacent images. More recently SAS systems have a single transmitter +projector and an array of receiving hydrophones so that the echoes received from any one ping can be reconstructed into a low-resolution image centred on the rotational axis of the real array. The images from the adjacent pings are cross-correlated to determine how the sonar has moved between pings. The sub-aperture (cf. map-drift algorithm) is thus determined by the extent of the physical array. From the cross-correlation, the differential sway can be estimated from a series of displaced single ping images just as in the map-drift algorithm. By measuring the rotational displacement between the images we can also estimate the relative yaw between single ping images. Using simulated, distorted SAS data as input, the proposed DPIA algorithm shows promise in that it can estimate yaw with sufficient accuracy to restore distorted SAS images very close to their diffraction limit. Here we present "before and after" simulated images from an extant SAS and demonstrate the power (and also the limitations) of the proposed DPIA technique.