High-Resolution CFISTA Algorithm for Forward-Looking Sonar Imaging

Abstract Multibeam forward-looking sonar is widely used in underwater target detection, submarine navigation and avoidance, seabed mapping, and other fields. The theoretical core of sonar imaging is conventional beamforming (CBF), whose performance is constrained by array aperture and frequency. At present, deconvolution beamforming algorithms can improve imaging resolution while maintaining good robustness. However, these power-spectrum-based deconvolution methods are not suitable for coherent targets, because they ignore the cross terms introduced by the magnitude-squared operation, which is inconsistent with the actual sonar imaging model. In response, a complex-domain fast iterative shrinkage-thresholding algorithm (CFISTA) is proposed to extend deconvolution beamforming to the complex domain and achieve high-resolution imaging of coherent targets, yielding narrower mainlobe and lower sidelobes. The experiments show that, at a signal-to-noise ratio (SNR) of −10 dB, the proposed method narrows the mainlobe width to less than one third that of CBF, while reducing sidelobe levels by about 15 dB. Further, CFISTA converges rapidly with modest computational cost, providing practical applications for coherent imaging with an improved azimuthal discrimination.