|本期目录/Table of Contents|

[1]赵重保,叶亭君,费斐,等.基于改进YOLOv4-tiny的安全标志检测*[J].中国安全生产科学技术,2025,21(6):149-158.[doi:10.11731/j.issn.1673-193x.2025.06.019]
　ZHAO Chongbao,YE Tingjun,FEI Fei,et al.Security flag detection based on improved YOLOv4-tiny[J].Journal of Safety Science and Technology,2025,21(6):149-158.[doi:10.11731/j.issn.1673-193x.2025.06.019]
点击复制

基于改进YOLOv4-tiny的安全标志检测*

《中国安全生产科学技术》[ISSN:1673-193X/CN:11-5335/TB]

卷:
21
期数:
2025年6期
页码:
149-158
栏目:
职业安全卫生管理与技术
出版日期:
2025-06-30

文章信息/Info

Title:
Security flag detection based on improved YOLOv4-tiny
文章编号:
1673-193X(2025)-06-0149-10
作者:
赵重保; 叶亭君; 费斐; 康士明; 赵雷; 王瑶涵; 宋泽阳
（1.西安科技大学 安全科学与工程学院，陕西 西安 710054；
2.华润资产管理有限公司，广东 深圳 518000；
3.华润欢乐颂商业管理有限公司，广东 深圳 518000；
4.山东省滨州市科技创新发展研究院，山东 滨州 256600）
Author(s):
ZHAO Chongbao;  YE Tingjun;  FEI Fei;  KANG Shiming;  ZHAO Lei;  WANG Yaohan;  SONG Zeyang
（1.College of Safety Science and Engineering,Xi’an University of Science and Technology,Xi’an Shaanxi 710054,China;
2.China Resources Asset Management Co.,Ltd.,Shenzhen Guangdong 518000,China;
3.China Resources Ode to Joy Business Management Co.,Ltd.,Shenzhen Guangdong 518000,China;
4.Binzhou Institute of Science and Technology Innovation and Development,Binzhou Shandong 256600,China）
关键词:
安全标志检测; 计算机视觉; YOLOv4-tiny; 注意力机制; Soft-NMS算法
Keywords:
safety signs detection;  computer vision;  YOLOv4-tiny;  attention mechanism;  Soft-NMS algorithm
分类号:
X913
DOI:
10.11731/j.issn.1673-193x.2025.06.019
文献标志码:
A
摘要:
为有效实现高效安全标志检测和对不安全行为预警，基于深度学习YOLOv4-tiny模型引入ECANet注意力机制，结合Soft-NMS算法提出1种用于检测安全标志的模型。模型中数据集包含2 000个安全标志，其中训练集1 620张、验证集180张和测试集200张。研究结果表明：该模型的检测精度达到97.76%，比YOLOv4-tiny和Faster RCNN卷积神经网络算法分别提高了7.55百分点和9.23百分点；改进的模型可避免YOLOv4-tiny和Faster RCNN卷积神经网络算法中出现的过拟合现象，泛化性能更好，在检测小目标区域和弱光条件下目标时，改进模型优势更加突出。研究结果可为施工场地安全标志的智能化监控与风险预警提供技术参考。
Abstract:
To effectively achieve efficient safety sign detection and unsafe behavior warning,a model for detecting safety signs is proposed by introducing the ECANet attention mechanism into the deep learning YOLOv4-tiny model and combining the Soft-NMS algorithm.The data set in the model contains 2 000 safety signs,including 1 620 training sets,180 verification sets and 200 test sets.Experimental results show that the detection accuracy of this model reaches 97.76%,which is 7.55% and 9.23% higher than that of YOLOv4-tiny and Faster RCNN respectively.In addition,the improved model avoids the overfitting phenomenon of YOLOv4-tiny and Faster RCNN,with better generalization performance.It also has more prominent detection advantages in small target areas and low light conditions.The research results can provide technical support for the intelligent monitoring and risk early warning of safety signs at construction sites.

参考文献/References:

［1］LECUN Y,BENGIO Y,HINTON G.Deep learning ［J］.Nature,2015,521:436-444.
［2］马庆禄,孙枭,唐小垚,等.基于改进YOLOv5s算法的隧道初期火灾检测模型 ［J］.中国安全科学学报,2023,33(10):214-223. MA Qinglu,SUN Xiao,TANG Xiaoyao,et al.Optimization method for tunnel initial fire detection based on YOLOv5s algorithm ［J］.China Safety Science Journal,2023,33(10):214-223.
［3］赵安新,黎梁,刘柯,等.煤矿井下钻孔裂隙识别技术 ［J］.西安科技大学学报,2023,43(6):1158-1167. ZHAO Anxin,LI Liang,LIU Ke,et al.Crack identification technology of bore in underground coal mine ［J］.Journal of Xi’an University of Science and Technology,2023,43(6):1158-1167.
［4］李华,薛曦澄,吴立舟,等.深度学习下吊装作业工人防护装备及吊钩检测方法 ［J］.安全与环境学报，2024，24（3）:1027-1035. LI Hua,XUE Xicheng,WU Lizhou,et al.Protective equipment and hook testing methods for workers in lifting operations under deep learning ［J］.Journal of Safety and Environment，2024，24（3）：1027-1035.
［5］XIAO Y,CHANG A,WANG Y,et al.Real-time object detection for substation security early-warning with deep neural network based on YOLO-V5［C］//2022 IEEE IAS Global Conference on Emerging Technologies (GlobConET).IEEE,2022:45-50.
［6］李禹纬,付锐,刘帆.改进YOLOv7的轻量化交通标 志检测算法［J］.太原理工大学学报,2024,55(1):195-203. LI Yuwei,FU Rui,LIU Fan.An improved lightweight traffic sign detection algorithm based on YOLOv7［J］.Journal of Taiyuan University of Technology,2024,55(1):195-203.
［7］李居朋,王颖慧,李刚.医学图像关键点检测深度学习方法研究与挑战 ［J］.电子学报,2022,50(1):226-237. LI Jupeng,WANG Yinghui,LI Gang.Research and challenges of medical image landmark detection based on deep learning ［J］.Acta Electronica Sinica,2022,50(1):226-237.
［8］陈禹,吴雪梅,张珍,等.基于改进YOLOv5s的自然环境下茶叶病害识别方法［J］.农业工程学报,2023,39(24):1-10. CHEN Yu,WU Xuemei,ZHANG Zhen,et al.Method for identifying tea diseases in natural environment using improved YOLOv5s ［J］.Transactions of the Chinese Society of Agricultural Engineering,2023,39(24):1-10.
［9］WANG C Y,BOCHKOVSKIY A,LIAO H Y M.YOLOv7:trainable bag-of-freebies sets new state-of-the-art for real-time object detectors［C］//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition,2023:7464-7475.
［10］FAROOQ J,MUAZ M,KHAN J K,et al.An improved YOLOv8 for foreign object debris detection with optimized architecture for small objects［J］.Multimedia Tools and Applications,2024,83(21):60921-60947.
［11］MARVELL Y A,GUNAWAN A A S,ISWANTO I A.Enhancing small object detection:development on car logo identification using the YOLO-NAS method for speed and accuracy improvements［J］.Procedia Computer Science,2024,245:528-537.
［12］WANG Z,LI B,LI W,et al.NAS-ASDet:an adaptive design method for surface defect detection network using neural architecture search［J］.Advanced Engineering Informatics,2024,61:102500.
［13］DING Y,CHEN X,ZHANG Y,et al.Smart building evacuation by tracking multi-camera network and explainable Re-identificationmodel［J］.Engineering Applications of Artificial Intelligence,2025,148:110394.
［14］韩江洪,卫星,陆阳,等.煤矿井下机车无人驾驶系统关键技术 ［J］.煤炭学报,2020,45(6):2104-2115. HAN Jianghong,WEI Xing,LU Yang,et al.Driverless technology of underground locomotive in coal mine ［J］.Journal of China Coal Society,2020,45(6):2104-2115.
［15］LEI M Y,ZHANG X H,YANG W J,et al.Research status and trends of visual pose detection and control for coal mine excavation equipment［J］.Journal of Coal Science and Engineering,2021,46(Supplement 2):1135-1148.
［16］LI H B,YUAN S,ZHOU W M,et al.The detection method for coal dust caused by chute discharge based on YOLOv4-tiny ［J］.Opto-Electronic Engineering,2021,48(6):210049-1-14.
［17］BOCHKOVSKIY A,WANG C Y,LIAO H M.YOLOv4:optimal speed and accuracy of object detection ［EB/OL］.（2020-04-23）［2024-09-29］.https://arxiv.org/pdf/2004.10934.
［18］GUO M H,XU T X.Attention mechanisms in computer vision:a survey［J］.Computational Visual Media,2022,8(3):331-368.
［19］WANG Q,WU B,ZHU P,et al.ECA-Net:efficient channel attention for deep convo-lutional neural networks［C］//Proceedings of the IEEE/CVF conference on computer visi-on and pattern recognition，2020:11534-11542.
［20］HU J,SHEN L,SUN G.Squeeze-and-excitation networks［C］//Proceedings of the IEEE conference on computer vision and pattern recognition，2018:7132-7141.
［21］LIN T Y,GOYAL P,GIRSHICK R,et al.Focal loss for dense object detection［C］//Proceedings of the IEEE international conference on computer vision，2017:2980-2988.
［22］GHIASI G,LIN T,LE Q.Dropblock:a regularization method for convolutional networks［EB/OL］.（2018-10-30）［2024-09-29］.10.48550/arXiv.1810.12890.
［23］ZHANG G,WANG C,XU B,et al.Three mechanisms of weight decay regularization［EB/OL］.（2018-10-29）［2024-09-29］.https://arxiv.org/pdf/1810.12281.
［24］EVERINGHAM M,ESLAMI S,GOOL L,et al.The pascal visual object classes challenge ［J］.International Journal of Computer Vision,2015,111:98-136.

相似文献/References:

[1]王治峰,杨涛,闵恒,等.基于深度学习模型的矿井火灾图像识别方法研究[J].中国安全生产科学技术,2025,21(8):146.[doi:10.11731/j.issn.1673-193x.2025.08.018]
　WANG Zhifeng,YANG Tao,MIN Heng,et al.Research on mine fire image recognition method based on deep learning model[J].Journal of Safety Science and Technology,2025,21(6):146.[doi:10.11731/j.issn.1673-193x.2025.08.018]

备注/Memo

备注/Memo:
收稿日期： 2024-09-29
* 基金项目： 陕西省高层次科技人才引进青年项目；新疆维吾尔自治区重点研发计划项目（2022B03025-2）
作者简介： 赵重保，硕士研究生，主要研究方向为能源与化工安全。
通信作者： 宋泽阳，博士，教授，主要研究方向为能源与化工安全、节能环保等。
更新日期/Last Update: 2025-07-01