FEV-YOLOv8n:轻量化安全帽佩戴检测方法  

FEV-YOLOv8n:Detection Methods for Wearing Lightweight Safety Helmets

作　　者：韩博 张婧婧[1,2,3] 鲁子翱[1,2,3] HAN Bo;ZHANG Jingjing;LU Ziao(College of Computer and Information Engineering,Xinjiang Agricultural University Urumqi 830052,China;Engineering Research Center of Intelligent Agriculture,Ministry of Education Urumqi 830052,China;Xinjiang Agricultural Informatization Engineering Technology Research Center,Urumqi 830052,China)

机构地区：[1]新疆农业大学计算机与信息工程学院,乌鲁木齐830052 [2]智能农业教育部工程研究中心,乌鲁木齐830052 [3]新疆农业信息化工程技术研究中心,乌鲁木齐830052

出　　处：《计算机测量与控制》2025年第1期69-77,84,共10页Computer Measurement &Control

基　　金：新疆维吾尔自治区自然科学基金资助项目(2022D01A202);新疆维吾尔自治区高校科研计划项目(XJEDU2020Y020);新疆维吾尔自治区研究生科研创新项目(XJ2024G124)。

摘　　要：针对基线YOLOv8n检测算法结构较复杂以及现有的安全帽佩戴检测算法参数量和计算量较大,难以在终端部署等问题,提出一种基于FEV-YOLOv8n的轻量化检测模型;设计一种轻量级的FasterC2f模块改进YOLOv8n的骨干网络,实现网络的参数量和计算量的降低;在FasterC2f模块中引入EMA注意力机制,融合空间依赖和位置信息,建立长短期的依赖关系,增强对目标表征的关注,以提高模型检测的精度;使用VoVGSCSP模块改进颈部网络,提高遮挡目标以及小目标的辨识度;实验结果表明,改进YOLOv8n模型map值为92.5%,相较于YOLOv8n算法,模型大小减少20%,计算量降低18.5%,参数量降低15.7%,为安全帽佩戴检测的轻量化研究提供理论参考。Aiming at the problems of more complicated structure for the baseline YOLOv8n detection algorithm and a large number of parameters and computational complexity for existing helmet wearing detection algorithms,it is difficult to be deployed at terminals,a lightweight detection model based on FEV-YOLOv8n is proposed.A lightweight FasterC2f module is designed to improve the backbone network of YOLOv8n,realizing the reduction of parameters and computation complexity of the network;the EMA attention mechanism is introduced into the FasterC2f module,fuses spatial dependence and positional information,establishes long and short-term dependence relationships,and enhances the attention to the target s representation,so as to improve the accuracy of the model s detection;and the VoVGSCSP is used to improve the neck network,and improve the recognition of occluded targets as well as small targets;Experimental results show that compared with the YOLOv8n algorithm,the improved YOLOv8n model reaches the map value by 92.5%,reduces the size by 20%,the computation by 18.5%,and the parameter quantity by 15.7%,which provides a theoretical reference for the lightweight of safety helmet wearing detection.

关 键 词：目标检测 安全帽 FasterC2f 轻量化 Efficient Multi-Scale Attention VoVGSCSP 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]