Beamforming for Active Sonar Imaging

This thesis addresses the use of adaptive beamforming and imaging methods applied to active sonar. While adaptive beamformers are known to offer improved performance when applied to passive sonars, their use is limited in active sonar systems. In this thesis, we investigate the potential performance gains of using adaptive beamforming methods for a range of active sonar applications. Key challenges that are addressed include coherent signals, arbitrary array geometries, computational load, and robustness. We also investigate the use of aperture coherence, or coherence over the elements of the receiver array, for improved sonar imaging. We conclude that in many cases, improved image quality is obtained by using adaptive beamforming methods. Noise, particularly directive noise such as sidelobe energy, is suppressed, edges of objects appear sharper and image resolution is improved. The minimum variance distortionless response (MVDR) beamformer offers superior resolution capabilities as well as the ability to suppress noise and interference, but requires robustification techniques in order to perform well in an active sonar setting. The amplitude and phase estimation (APES) beamformer is robust and displays similar noise and interference suppression capabilities as the MVDR, but somewhat lower resolution. The low complexity adaptive (LCA) beamformer offers much of the same improvements as the fully adaptive MVDR at a fraction of the computational cost, is robust, and is particularly well suited for small arrays. The coherence factor and the closely related scaled Wiener postfilter represent a promising class of adaptive imaging methods that work in a fundamentally different way than the MVDR and APES beamformers. Based on the concept that noise and interference appear more random across the receiver array than the echo from a reflecting object, they effectively attenuate noisy image pixels in an adaptive way, resulting in images which appear less contaminated by noise. As a result, shadow regions are enhanced and details in the image appear more clearly defined.