Underwater target detection based on wavelet transform and attention mechanism
Abstract
针对侧扫声呐图像在水下混响作用下易受噪声干扰、导致目标检测性能受限的问题，提出了一种融合小波变换和注意力机制的水下目标检测模型。该模型利用小波变换实现高、低频特征分解，并结合注意力机制对低频全局结构与高频局部细节进行差异化增强，有效抑制噪声干扰，突出目标特征表达。实验结果表明，所提模型在mAP@0.5和mAP@0.5:0.95上分别达到87.0%和56.9%，参数量仅为2.37 M，FLOPs为5.99 G，展现出高精度与轻量化的综合优势，可为深海矿产勘探和水下考古等任务提供可靠的技术支撑。
Fields of Study & Subjects
Underwater Acoustics Research
Advanced Neural Network Applications
Underwater Vehicles and Communication Systems
Publication details
- Journal / source
- Natural Science of Hainan University
- Volume
- 44
- Issue
- 2
- Pages
- 174–174
- Publication date
- 1 January 2026
- Language
- ZH