To address the limitation of target detection performance in sidescan sonar images caused by noise interference from underwater reverberation, this paper proposes an underwater target detection model that integrates wavelet transform and an attention mechanism. The model employs wavelet transform to decompose high- and low-frequency features, and the attention mechanism adaptively enhances low-frequency global structures and high-frequency local details, effectively suppressing noise interference and highlighting target feature representations. Experimental results demonstrate that the proposed model achieves mAP@0.5 and mAP@0.5:0.95 scores of 87.0% and 56.9%, respectively, with only 2.37×10⁶ parameters and 5.99 G FLOPs, achieving a favorable balance between high accuracy and lightweight design. The proposed model provides reliable technical support for tasks such as deep-sea mineral exploration and underwater archaeology.