基于YOLOv8n的轻量化鱼类检测算法  

Lightweight fish detection algorithm based on YOLOv8n

作　　者：王明慧 陈燕[1,2,3] 寇立伟 窦银科[1,2,3] WANG Minghui;CHEN Yan;KOU Liwei;DOU Yinke(College of Electrical and Power Engineering,Taiyuan University of Technology,Taiyuan 030024,China;Shanxi Energy Internet Research Institute,Taiyuan 030032,China;Key Laboratory of Cleaner Intelligent Control on Coal and Electricity,Ministry of Education,Taiyuan 030024,China)

机构地区：[1]太原理工大学电气与动力工程学院,山西太原030024 [2]山西省能源互联网研究院,山西太原030032 [3]煤电清洁智能控制教育部重点实验室,山西太原030024

出　　处：《现代电子技术》2025年第5期79-85,共7页Modern Electronics Technique

基　　金：国家重点研发计划(2022YFC2807603);山西省基础研究计划资助项目(202203021211175,202103021223048)。

摘　　要：目前极区是通过声呐探测结合人工捕捞统计的方法对鱼类进行监测考察,该方法受经济成本和作业区域、时间的限制。基于深度学习的目标检测算法可在满足经济性要求的同时对鱼类进行识别检测,然而传统的目标检测算法往往参数量、计算量较大,无法适应极区能耗、存储受限的严苛条件。针对这一问题,文中提出一种改进YOLOv8n的轻量化鱼类检测算法,在骨干和颈部网络中使用GhostC2f代替C2f,用GhostConv代替网络中部分Conv;在骨干网络中引入EMA注意力机制,以提高特征提取能力;最后,使用计算过程更简单的MPDIoU损失函数代替CIoU,提高检测速度。在自制的鱼类数据集上实验表明:改进后的算法参数量和计算量分别变为1.49×10^(6)和4.7×10^(9),仅用了原YOLOv8n算法49.67%的参数实现了略优于YOLOv8n的检测精度;部署到嵌入式设备Jetson Xavier NX中检测速度能达到47 f/s,可以为硬件条件受限情况下的鱼类检测提供技术支撑。At present,the fishes in polar region are monitored and investigated by combining sonar detection with artificial fishing statistics.However,this method is limited by economic cost,operation area and time.Object detection algorithms based on deep learning can identify and detect fishes while meeting economic requirements.In the traditional object detection algorithms,however,there are a large number of parameters and large calculation quantity,so the algorithms fail to adapt to the harsh conditions of energy consumption and storage limitations in the polar region.In view of this,an improved lightweight fish detection algorithm based on YOLOv8n is proposed.In the algorithm,the GhostC2f module is used to replace C2f in the backbone and neck networks,and GhostConv is used to replace part of the Conv in the network,and the EMA is introduced in the backbone network to improve the feature extraction ability.Finally,the loss function MPDIoU,which has a simpler calculation,is used to replace the CIoU to improve the detection speed.Experiments on the self-made fish dataset show that the number of parameters and computation burden of the improved algorithm become 1.49×10^(6) and 4.7×10^(9),respectively,and only 49.67%of the parameters of the original YOLOv8n are used to achieve a detection accuracy slightly higher than that of the YOLOv8n.When the proposed algorithm is deployed in the embedded Jetson Xavier NX,it can achieve an inspection speed of up to 47 f/s,so it can provide technical support for fish detection in hardware-constrained situations.

关 键 词：鱼类检测 YOLOv8n 轻量化 极区 声呐探测 EMA注意力机制 

分 类 号：TN911-34[电子电信—通信与信息系统] S951.2[电子电信—信息与通信工程]