农业装备与车辆工程2025，Vol.63Issue(6)：103-108,118,7.DOI:10.3969/j.issn.1673-3142.2025.06.019
基于ACl-YOLOv8s的工厂吸烟行为实时检测
Real-time detection of smoking behavior in factories based on ACI-YOLOv8s
王泓博1
作者信息
- 1. 上海烟草机械有限责任公司,上海 201206
- 折叠
摘要
Abstract
With the continuous improvement of safety management requirements in factories,real-time monitoring and prevention of fire and other safety hazards have become an urgent problem.As an important cause of fire,the detection and management of smoking behavior was of great significance.Aiming at the problems of omission,misjudgment and insufficient recognition of small targets.In traditional detection methods,an improved smoking behavior detection model based on YOLOv8s,ACI-YOLOv8s was proposed,which incorporated an asymptotic feature pyramid to enhance the combination of low-level details and high-level semantic information,and a CBAM attention module embedded in the backbone network to dynamically optimized the feature weight allocation.In addition,the accuracy and convergence speed of target localization were further improved by improving the regression loss function.The experimental results showed that the ACI-YOLOv8s model achieved improvements of 1.5%,0.9%,and 3.3%in precision,recall,and detection speed respectively compared with the baseline YOLOv8s model,which could provide technical support for intelligent security monitoring in factories.关键词
工厂/吸烟/ACI-YOLOv8s/小目标检测/深度学习Key words
factories/smoking/ACI-YOLOv8s/small target detection/deep learning分类
信息技术与安全科学引用本文复制引用
王泓博..基于ACl-YOLOv8s的工厂吸烟行为实时检测[J].农业装备与车辆工程,2025,63(6):103-108,118,7.
