Feature-Enhanced Beamforming for Underwater 3-D Acoustic Imaging

A planar array of sensors and beamforming algorithms are required for underwater 3-D acoustic imaging system. However, due to limited aperture and sensor numbers, conventional beamforming usually results in wide mainlobes and high sidelobes, which severely degrade the image quality. To date, using unconventional beamforming algorithms to improve the imaging quality of underwater 3-D acoustic systems have been rarely investigated. This article proposes a feature-enhanced beamforming (FEBF) algorithm, with the goal of improving the imaging quality by utilizing more appropriate prior knowledge that objects in the angular domain in underwater 3-D imaging scenes are generally sparse but locally dense. The proposed FEBF employs the $l_{1}$-total variation (L1-TV) mixed norm regularization. The penalized beamforming optimization is reformed into convex programming with linear constraints and solved by alternating direction method of multipliers. The results of simulation and experiment demonstrate that the proposed FEBF has the ability to effectively improve the image quality when compared with the other existing beamforming methods. This study also proposes a 2-D fast Fourier transform based acceleration strategy, which effectively reduces the computational time of the FEBF algorithm by about 4.3 x 10(3) times for forming 200 x 200 beams. The accelerated version of the proposed FEBF method is considered to have potential for real-time imaging systems.