Abstract: Aiming at the problem of the decline of target detection capability caused by broadband background noise interference in underwater vehicles and other platforms, a deep learning broadband background noise suppression method suitable for multi-channel hydrophone arrays is proposed. This method preserves the phase information by cascading the multi-channel frequency domain characteristics of the interference signal, and a learning model is established to estimate the spectral characteristics of pure target signals by using a deep complex neural network to achieve interference suppression, and then the conventional beamforming method is used to achieve object detection and tracking. The simulation results show that under the dual target signal conditions of −15 dB, −20 dB and −25 dB signal-to-interference ratio, the proposed method can effectively reduce the influence of near-field interference and improve the detection ability of conventional beamforming. The lake test results show that the proposed method can adaptively realize the suppression of platform broadband background noise interference, and effectively improve the target detection and tracking performance.