Reconstruction of Under-sampled SAS Data Using the WIPE Algorithm

Abstract

Synthetic aperture sonar (SAS) processing provides seafloor images with higher resolution than conventional sonar processing. The ability to achieve large area coverage rate (ACR) is a key benefit of SAS for mine countermeasures (MCM) applications. One bound on ACR for SAS systems is that, in order to ensure optimal image quality, the sonar platform should be towed at no more than half the length of its hydrophone array per ping. When this speed limit is violated, the resulting SAS image is degraded with undersampling artifacts. In this work, we evaluated potential iterative image deconvolution techniques in order to increase ACR in high frequency MCM for use in the Autonomous Operations Future Naval Capabilities (AO FNC). In particular, we implemented and evaluated the WIPE iterative deconvolution algorithm for reconstructing SAS images from data that have been undersampled due to platform speeds in excess of the SAS speed limit. WIPE is a regularized form of CLEAN that trades resolution for inversion stability. We have successfully implemented one-and two-dimensional versions of the WIPE algorithm. We applied the WIPE algorithm to simulated scenes as well as collected data scenes that were down-sampled for demonstration purposes. We found that the WIPE algorithm produced consistently better images than the previously implemented CLEAN algorithm and often better, for the bright objects, than the under-sampled images. However, while WIPE provides an improvement over CLEAN for under-sampled data, WIPE does not satisfactorily reconstruct the shadow contrast in the images, which is a major classification feature for high frequency MCM. This is a consequence of the development of WIPE for radio astronomy using Fourier synthesis techniques for image formation. Due to the poor shadow contrast reconstruction, we conclude that WIPE is more appropriate for anti-submarine warfare (ASW) SAS applications where this characteristic is not currently used in classification techniques.