A TRPCA-based reverberation suppression method for detecting slowly-moving small underwater targets in very shallow water environments

This paper proposes a tensor robust principal component analysis (TRPCA)-based reverberation suppression method for detecting slowly-moving small underwater targets in very shallow water environments. By directly processing the three-dimensional (bearing × range × ping) sonar data without destroying the tensor structure, the proposed method effectively decomposes three-dimensional sonar data into a low-rank component (mainly includes the bottom reverberation) and a sparse component. The majority of the bottom reverberation can be suppressed by removing the low-rank component, and the sparse component is retained where the target highlights are contained. Simulation results under low signal-to-reverberation ratio (−20 dB) conditions show that TRPCA-alternating direction method of multipliers outperforms conventional RPCA and frequency-filtered TRPCA (TRPCA-FTNN) methods, improving the bias of integral sidelobe ratio by about 1 dB–9 dB. Lake experiments in very shallow water (2–3 m depth) further validate the method’s practical capability of suppressing strong reverberation and enhancing the target detection performance.