|本期目录/Table of Contents|

[1]傅展斌,李俚,甘进浩,等.融合关键点检测的制造车间工人不安全行为识别算法研究*[J].中国安全生产科学技术,2026,22(4):155-162.[doi:10.11731/j.issn.1673-193x.2026.04.019]
　FU Zhanbin,LI Li,GAN Jinhao,et al.An integrated keypoint detection algorithm for identifying unsafe behaviors of workers in manufacturing workshops[J].Journal of Safety Science and Technology,2026,22(4):155-162.[doi:10.11731/j.issn.1673-193x.2026.04.019]
点击复制

融合关键点检测的制造车间工人不安全行为识别算法研究*

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

卷:
22
期数:
2026年4期
页码:
155-162
栏目:
职业健康与防护
出版日期:
2026-04-30

文章信息/Info

Title:
An integrated keypoint detection algorithm for identifying unsafe behaviors of workers in manufacturing workshops
文章编号:
1673-193X(2026)-04-0155-08
作者:
傅展斌; 李俚; 甘进浩; 黄珠芹
（1.广西大学 机械工程学院，广西 南宁530004；
2.广西美斯达集团有限公司，广西 南宁 530009)
Author(s):
FU Zhanbin;  LI Li;  GAN Jinhao;  HUANG Zhuqin
（1.School of Mechanical Engineering,Guangxi University,Nanning Guangxi 530004,China;
2.Guangxi Meisida Group Co.,Ltd.,Nanning Guangxi 530009,China）
关键词:
区域入侵; 安全帽检测; 不安全行为; 射线法; 关键点检测
Keywords:
area intrusion;  safety helmet detection;  unsafe behavior;  ray-casting method;  keypoint detection
分类号:
X913
DOI:
10.11731/j.issn.1673-193x.2026.04.019
文献标志码:
A
摘要:
为解决厂区车间工人安全帽佩戴检测与危险区域误入识别中因目标多尺度、作业环境复杂及人员姿态变化导致的漏检与准确率不足的问题，提出1种基于改进YOLOv8n-Pose的综合识别算法KP-YOLOv8。该方法包含目标检测与行为判断2个部分：在目标检测部分，在主干网络引入高低频特征自适应增强（high and low frequency features adaptive enhancement,HLFAE）模块以增强多尺度特征感知能力，在颈部结构采用DySample上采样以改善特征对齐与细节恢复，在检测头输入端嵌入深度分组全局坐标注意力(depth grouping global coordinate attention,DGGCA)以加强全局信息利用与小目标定位；在行为判断部分，引入射线法并设计两点射线法，结合提取的关键点信息实现2类不安全行为的精准判定。研究结果表明：KP-YOLOv8相比同类算法具有更高检测精度，较基准模型mAP与kp-mAP分别提升3.0与3.4百分点，能够有效识别2类不安全行为。研究结果可为提升工厂车间安全管理的智能化水平提供参考。
Abstract:
In order to solve the problem of missed detection and insufficient accuracy in hard-hat wearing detection and hazardous-area intrusion recognition for workshop workers in factory plants,which are caused by multi-scale targets,complex working environments,and changes in worker posture,this paper proposes KP-YOLOv8,a comprehensive recognition algorithm based on an improved YOLOv8n-Pose.The method consists of two parts,namely object detection and behavior judgment.In the object detection part,the high and low frequency features adaptive enhancement （HLFAE） module is introduced into the backbone network to enhance multi-scale feature perception,DySample upsampling is adopted in the neck structure to improve feature alignment and detail recovery,and depth grouping global coordinate attention （DGGCA） is embedded at the input end of the detection head to strengthen global information utilization and small-object localization.In the behavior judgment part,a ray-casting method is introduced and a two-point ray-casting method is designed.By combining the extracted keypoint information,the method achieves accurate judgment of two types of unsafe behaviors.The results show that KP-YOLOv8 has higher detection accuracy than similar algorithms.Compared with the baseline model,mAP and kp-mAP increase by 3.0 and 3.4 percentage points,respectively,and the method can effectively recognize two types of unsafe behaviors.The results provide a reference for improving the intelligence level of safety management in factory workshops.

参考文献/References:

［1］WANG H,LYU L,LI X,et al.A safety managem-ent approach for Industry 5.0’s human-centered m-anufacturing based on digital twin［J］.Journal of Manufacturing Systems,2023,66:1-12.
［2］HAN L,FENG H,LIU G,et al.A real-time intelligent monitoring method for indoor evacuee distribution based on deep learning and spatial division［J］.Journal of Building Engineering,2024,92:109764.
［3］罗国富,王源,李浩,等.基于改进YOLOv5s的智能车间工人不安全行为实时检测方法［J］.计算机集成制造系统,2024,30(5):1610-1619. LUO Guofu,WANG Yuan,LI Hao,et al.Real-time detection method of unsafe behavior of workers inintelligent workshop based on improved YOLOv5s［J］.Computer Integrated Manufacturing Systems,2024,30 (5):1610-1619.
［4］SONG C,LI Y.A YOLOv8 algorithm for safety helmet wearing detection in complex environment［J］.Scientific Reports,2025,15(1):24 236.
［5］MEI X,ZHOU X,XU F,et al.Human intrusion detection in static hazardous areas at construction sites:deep learning-based method［J］.Journal of Construction Engineering and Management,2023,149(1):04022142.
［6］YANG T,GUO Y,LI D,et al.Vision-Based obstacle detection in dangerous region of coal mine driverless rail electric locomotives［J］.Measurement,2025,239:115514.
［7］田枫,白欣宇,刘芳,等.1种基于视频的油田危险区域入侵检测智能综合识别技术研究［J］.中国安全生产科学技术,2022,18(3):68-75. TIAN Feng,BAI Xinyu,LIU Fang,et al.Research on a video-based intelligent integrated identification technology for oilfield dangerous area intrusion detection［J］.Journal of Safety Science and Technology,2022,18(3):68-75.
［8］WANG Q,WU B,ZHU P,et al.ECA-Net:efficient channel attention for deep convolutional neural networks［C］//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2020:11534-11542.
［9］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.
［10］LIU W,LU H,FU H,et al.Learning to upsample by learning to sample［C］//Proceedings of the IEEE/CVF International Conference on Computer Vision.2023:6027-6037.
［11］HOU Q,ZHOU D,FENG J.Coordinate attention for efficient mobile network design［C］//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2021:13713-13722.
［12］蒋新源.基于Transformer的矿场掘进面危险区域入侵检测系统的设计与实现［D］.北京:北京邮电大学,2024.
［13］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.
［14］HE K,GKIOXARI G,DOLLAR P,et al.Mask r-cnn［C］//Proceedings of the IEEE International Conference on Computer Vision.2017:2961-2969.
［15］KHANAM R,HUSSAIN M.YOLOv11:an overview of the key architectural enhancements［EB/OL］.(2024-10-23)［2025-12-13］.https://doi.org/10.48550/ar-Xiv.2410.17725.
［16］YANG J,ZENG A,LIU S,et al.Explicit box detection unifies end-to-end multi-person pose estimation［EB/OL］.(2023-02-03)［2025-12-13］.https://doi.org/10.48550/arXiv.2302.01593.
［17］LU P,JIANG T,LI Y,et al.Rtmo:towards high-performance one-stage real-time multi-person pose estimation［C］//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2024:1491-1500.
［18］刘泽锋,韩豫,李文涛,等.视觉联合交集下施工危险区域侵入行为动态感知方法［J］.安全与环境工程,2023,30 (4):18-25. LIU Zefeng,HAN Yu,LI Wentao,et al.Dynamic perception method of intrusion behavior in construction dangerous area under visual joint intersection［J］.Safety and Environmental Engineering,2023,30(4):18-25.

相似文献/References:

[1]郭师虹,井锦瑞,张潇丹,等.基于改进的YOLOv4安全帽佩戴检测研究[J].中国安全生产科学技术,2021,17(12):135.[doi:10.11731/j.issn.1673-193x.2021.12.022]
　GUO Shihong,JING Jinrui,ZHANG Xiaodan,et al.Research on detection of safety helmet wearing based on improved YOLOv4[J].Journal of Safety Science and Technology,2021,17(4):135.[doi:10.11731/j.issn.1673-193x.2021.12.022]
[2]田枫,白欣宇,刘芳,等.1种基于视频的油田危险区域入侵检测智能综合识别技术研究*[J].中国安全生产科学技术,2022,18(3):68.[doi:10.11731/j.issn.1673-193x.2022.03.010]
　TIAN Feng,BAI Xinyu,LIU Fang,et al.Research on intelligent comprehensive recognition technology of intrusion detection in oilfield dangerous area based on video[J].Journal of Safety Science and Technology,2022,18(4):68.[doi:10.11731/j.issn.1673-193x.2022.03.010]
[3]程磊,张俊展,景国勋,等.基于YOLOv8的煤矿安全帽和安全背心检测算法*[J].中国安全生产科学技术,2025,21(2):115.[doi:10.11731/j.issn.1673-193x.2025.02.015]
　CHENG Lei,ZHANG Junzhan,JING Guoxun,et al.Detection algorithm of safety hat and safety vest in coal mine based on YOLOv8[J].Journal of Safety Science and Technology,2025,21(4):115.[doi:10.11731/j.issn.1673-193x.2025.02.015]

备注/Memo

备注/Memo:
收稿日期： 2025-12-13；网络首发日期： 2026-04-08
* 基金项目： 南宁市揭榜制科技项目(20221241)
作者简介： 傅展斌，硕士研究生，主要研究方向为计算机视觉的工业人体行为识别。
通信作者： 李俚，硕士，教授，主要研究方向为行为安全、安全管理。
更新日期/Last Update: 2026-04-29