| 970 | 13 | 457 |
| 下载次数 | 被引频次 | 阅读次数 |
不安全行为是建筑工人自身安全的最主要威胁，为准确刻画建筑工人不安全行为并揭示个体间的差异化特征，提出了一种基于姿态特征的建筑工人不安全行为刻画方法。该方法根据现场采集的建筑工人施工视频数据，利用OpenPose卷积神经网络提取建筑工人姿态的关键点，并结合用户画像理论和潜在类别聚类分析方法确定反映建筑工人行为特征的事实标签和模型标签。结果表明：以高处作业人员为例，该方法在实际施工视频测试中可取得95.60%的平均准确率和87.02%的平均精确率，且将建筑工人划分为5种不同的行为偏好群体是合理的。该研究结果对揭示建筑工人个体特征差异、促进建筑工人行为的个性化安全管理具有重要的现实意义。
Abstract：Construction workers' unsafe behaviors seriously threaten their safety and health.To describe unsafe behaviors accurately and reveal the different characteristics of behaviors among workers, this study presents a description method of construction workers' unsafe behaviors based on posture characteristics.The proposed method uses OpenPose deep convolutional network to extract key points of workers based on the collected on-site construction videos, then adopts the user profile theory and the latent class cluster analysis method to determine the fact label and model label reflecting the behavior characteristics of construction workers.Taking high-altitude operation workers as an example, the results show that the proposed method achieves an average accuracy of 95.60% and an average precision of 87.02% in the test of actual construction video, and it is reasonable to divide construction workers into 5 different behavior prefe-rence groups.The research has practical significance to reveal the differences of individual characteristics of construction workers and promote the personalized safety management of workers' behaviors.
[1] Suraji A,Royduff A,Jpeckitt S.Development of causal model of construction accident causation[J].Journal of Construction Engineering and Management,2001,127(4):337-344.
[2] Tang N,Hu H,Xu F,et al.Personalized safety instruction system for construction site based on internet technology[J].Safety Science,2019,116:161-169.
[3] 康海燕，王紫豪，于爱民，等.基于网络日志的用户行为刻画与预测研究[J].郑州大学学报(理学版),2019,51(3):48-54.
[4] 高明，金澈清，钱卫宁，等.面向微博系统的实时个性化推荐[J].计算机学报，2014,37(4):963-975.
[5] Tang N,Hu H,Xu F,et al.A personalized Human Factors Ana-lysis and Classification System (HFACS) for construction safety management based on context-aware technology[J].Enterprise Information Systems,2022,16(1):141-166.
[6] Oswald D,Sherratt F,Smith S.Problems with safety observation reporting:A construction industry case study[J].Safety Science,2018,107:35-45.
[7] Kim K,Kcho Y.Effective inertial sensor quantity and locations on a body for deep learning-based worker's motion recognition[J].Automation in Construction,2020,113:103126.DOI:10.1016/j.autcon.2020.103126.
[8] 张明媛，曹天卓，赵雪峰.基于ANN识别施工人员跌落险兆事故的研究[J].安全与环境学报，2018,18(5):1703-1710.
[9] 张博，宋元斌，熊若鑫，等.融合人体关节点的安全帽佩戴检测[J].中国安全科学学报，2020,30(2):177-182.
[10]王雨生，顾玉宛，封晓晨，等.基于姿态估计的安全帽佩戴检测方法研究[J].计算机应用研究，2021,38(3):937-940.
[11]Guo H L,Yu Y T,Ding Q H,et al.Image-and-skeleton-based parameterized approach to real-time identification of construction worker' unsafe behaviors[J].Journal of Construction Enginee-ring and Management,2018,144(6):4018042.DOI:10.1061/(ASCE)CO.1943-7862.0001497.
[12]Ding L Y,Fang W L,Luo H B,et al.A deep hybrid learning model to detect unsafe behavior:Integrating convolution neural networks and long short-term memory[J].Automation in Construction,2018,86:118-124.
[13]Cao Z,Tomas S,Wei S,et al.Realtime multi-person 2d pose estimation using part affinity fields[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR).Hondulu,HI,USA:IEEE,2017:7291-7299.
[14]Tomas S,Hanbyul J,Iain M,et al.Hand keypoint detection in single images using multiview bootstrapping[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Re-cognition (CVPR).Hondulu,HI,USA:IEEE,2017:1145-1153.
[15]Washington State Department of Labor and Industries.Ergonomics (WAC 296-62-051) [EB/OL].(2003-04-11) [2021-05-21].https://www.humanics-es.com/ergorulewithappendices.pdf.
[16]韩锟，黄泽帆.基于人体姿态动态特征的跌倒行为识别方法[J].湖南大学学报(自然科学版),2020,47(12):69-76.
[17]Ayhan B U,Tokdemir O B.Accident analysis for construction safety using latent class clustering and artificial neural networks[J].Journal of Construction Engineering and Management,2020,146(3):4019114.DOI:10/1061/(ASCE)CO.1943-7862.0001762.
基本信息:
DOI：10.13578/j.cnki.issn.1671-1556.20210652
中图分类号:TU714;TP183;TP391.41
引用信息:
[1]段品生,周建亮.基于姿态特征的建筑工人不安全行为刻画方法[J].安全与环境工程,2022,29(03):1-8.DOI:10.13578/j.cnki.issn.1671-1556.20210652.
基金信息:
国家自然科学基金项目(72171224); 教育部人文社科规划基金项目(19YJAZH122); 江苏省研究生科研与实践创新计划项目(KYCX21_2478); 中国矿业大学实验室开放基金项目
2022-05-27
2022-05-27