通信与信息技术Issue(3)：125-128,133,5.
基于YOLOv5的水下小目标生物检测方法研究
Biological detection method based on YOLOv5
梁云龙 1王鸿鹏 1蒋强1
作者信息
- 1. 沈阳理工大学 自动化与电气工程学院,沈阳 110159
- 折叠
摘要
Abstract
In the actual fishing,underwater organisms are located in a complex environment and numerous debris on the seabed,which is difficult to identify accurately.In view of this,a method of underwater small target biometry and detection is proposed.The algo-rithm takes YOLOv5 as the model,changing the CloU to EloU for the problem that w and h of pictures cannot be in or out at the same time;for the interference problem of useless information,SENet attention mechanism is introduced to create the c3se module;improve the ability of the model to identify objects at different scales,and create SPPFCSPC structure at multi-scale links to improve the small target detection effect of the model.The ablation experiment of small target detection was performed through an underwater biological data set of 2920 pictures of data collection in multiple scenarios.The experimental results show that compared with the YOLOv5 algorithm,the three improvements improve the accuracy by 0.4%,3.3%and 4.2%,and the overall accuracy by 4.9%.关键词
YOLOv5/注意力机制/SPPFCSPC/小目标检测Key words
YOLOv5/Attention mechanism/SPPFCSPC/Small target detection分类
信息技术与安全科学引用本文复制引用
梁云龙,王鸿鹏,蒋强..基于YOLOv5的水下小目标生物检测方法研究[J].通信与信息技术,2025,(3):125-128,133,5.
