福建电脑2025，Vol.41Issue(9)：11-15,5.DOI:10.16707/j.cnki.fjpc.2025.09.003
轻量型鱼类实时检测模型研究
A Lightweight Real-Time Fish Detection Model
陈德金 1冯德旺1
作者信息
- 1. 福建农林大学计算机与信息学院 福州 350000
- 折叠
摘要
Abstract
To address the issues of poor real-time performance and high computational resource consumption in underwater fish detection,this paper proposes a lightweight fish detection model based on an improved YOLOv9 algorithm.Build a high-definition dataset containing 88 freshwater fish species through multi angle shooting,and enhance image quality by combining dehazing algorithms.Introducing ELA attention mechanism into the model and optimizing the feature extraction module significantly improves detection performance.The experimental results show that the improved model has a single frame detection time of only 14.5ms,an average accuracy of 91%,which is 4%higher than the original YOLOv9,and a reduced parameter size of 1.9M.It is suitable for real-time detection of embedded devices and provides effective technical support for fishery resource management.关键词
鱼类检测/注意力机制/YOLOv9算法/图像增强Key words
Fish Detection/Attention Mechanism/YOLOv9/Image Enhancement分类
信息技术与安全科学引用本文复制引用
陈德金,冯德旺..轻量型鱼类实时检测模型研究[J].福建电脑,2025,41(9):11-15,5.
