铁路工人人体行为识别模型

doi:10.16265/j.cnki.issn1003-3033.2022.06.2179

中国安全科学学报 ›› 2022, Vol. 32 ›› Issue (6): 17-22.doi: 10.16265/j.cnki.issn1003-3033.2022.06.2179

• 安全社会科学与安全管理 • 上一篇下一篇

铁路工人人体行为识别模型

黄珍珍¹^,²(), 肖硕¹^,^**(), 王钰³, 陈伟¹, 王升志¹, 江海峰¹

1 中国矿业大学计算机科学与技术学院,江苏徐州 221116
2 中国矿业大学图书馆,江苏徐州 221116
3 卡斯柯信号有限公司,北京 100070

收稿日期:2022-01-12 修回日期:2022-04-14 出版日期:2022-06-28 发布日期:2022-12-28
通讯作者: 肖硕
作者简介:
黄珍珍 (1981—),女,江苏徐州人,博士研究生,副研究馆员,研究方向为人体行为识别、物联网技术、大数据分析。E-mail: huangzhenzhen@cumt.edu.cn。
肖硕,副教授
王钰,高级工程师
陈伟,教授
江海峰,副教授
基金资助:
国家自然科学基金资助(62071470); 国家自然科学基金资助(61971421); 国家自然科学基金资助(51874300); 徐州市科技计划项目(KC20167)

Human activity recognition model of railway workers

HUANG Zhenzhen¹^,²(), XIAO Shuo¹^,^**(), WANG Yu³, CHEN Wei¹, WANG Shengzhi¹, JIANG Haifeng¹

1 School of Computer Science and Technology, China University of Mining and Technology, Xuzhou Jiangsu 221116, China
2 Library, China University of Mining and Technology, Xuzhou Jiangsu 221116, China
3 Casco Signal Company Limited, Beijing 100070, China

Received:2022-01-12 Revised:2022-04-14 Online:2022-06-28 Published:2022-12-28
Contact: XIAO Shuo

摘要/Abstract

摘要：

为提高铁路工人施工安全系数,采用基于人体行为识别(HAR)的智能化监测方法,估计铁路工人在施工过程中的动作;使用端到端自动提取数据特征的深度学习方法搭建网络,提高行为识别精度和模型泛化性;鉴于循环神经网络并行能力差,收敛时间长,提出结合空洞卷积与自注意力机制的深度学习模型;使用WISDM和MobiAct公开数据集,分别识别2个数据集上的基本动作和跌倒、撞击等行为。结果表明:相比于卷积神经网络(CNN)、长短期记忆(LSTM)网络、深度卷积LSTM网络,该模型具有更好的识别精度和性能,能够实现更准确的工人行为划分。

关键词: 铁路工人, 人体行为识别(HAR), 深度学习, 空洞卷积, 自注意力机制

Abstract:

In order to improve the construction safety factor of railway workers, the intelligent monitoring method based on HAR was used to estimate the action of railway workers in the construction process. The deep learning method of end-to-end automatic extraction of data features is applied to build a network to improve the accuracy of behavior recognition and model generalization. In view of the poor parallel ability and long convergence time of the recurrent neural network(CNN), a deep learning model combining cavity convolution and self-attention mechanism is proposed. The WISDM and MobiAct public datasets are used to identify the basic actions and fall and impact behaviors on the two datasets. The results show that compared with convolutional neural network(CNN), long-term and short-term memory(LSTM) network and deep convolutional LSTM neural network, the model has better recognition accuracy and performance, and can realize more accurate division of worker behavior.

Key words: railway workers, human activity recognition(HAR), deep learning, dilated convolution, self-attention mechanism

黄珍珍, 肖硕, 王钰, 陈伟, 王升志, 江海峰. 铁路工人人体行为识别模型[J]. 中国安全科学学报, 2022, 32(6): 17-22.

HUANG Zhenzhen, XIAO Shuo, WANG Yu, CHEN Wei, WANG Shengzhi, JIANG Haifeng. Human activity recognition model of railway workers[J]. China Safety Science Journal, 2022, 32(6): 17-22.

图/表 8

图1

图2

表1

表2

表3

表4

图3

图4

参考文献 13

[1]	晋良海, 张荣坤, 郑霞忠, 等. 隧洞钻孔作业工效及低位作业姿势研究[J]. 中国安全科学学报, 2016, 26(4): 66-71.
	JIN Lianghai, ZHANG Rongkun, ZHENG Xiazhong, et al. Research on simulation of drilling work posture and optimization of lowing work posture for tunnel[J]. China Safety Science Journal, 2016, 26(4): 66-71.
[2]	熊若鑫, 宋元斌, 王宇轩, 等. 基于CNN的3D姿势估计在建筑工人行为分析中的应用[J]. 中国安全科学学报, 2019, 29(7): 64-69. doi: 10.16265/j.cnki.issn1003-3033.2019.07.011
	XIONG Ruoxin, SONG Yuanbin, WANG Yuxuan, et al. Application of convolutional neural network-based 3D posture estimation in behavioral analysis of construction workers[J]. China Safety Science Journal, 2019, 29(7): 64-69. doi: 10.16265/j.cnki.issn1003-3033.2019.07.011
[3]	强茂山, 张东成, 江汉臣. 基于加速度传感器的建筑工人施工行为识别方法[J]. 清华大学学报:自然科学版, 2017, 57(12): 1338-1344.
	QIANG Maoshan, ZHANG Dongcheng, JIANG Hanchen. Recognizing construction worker activities based on accelerometers[J]. Journal of Tsinghua University:Science and Technology, 2017, 57(12): 1338-1344.
[4]	杜成飞. 基于机器学习的铁路工务人员行为识别方法[J]. 计算机系统应用, 2019, 28(7): 199-205.
	DU Chengfei. Railway engineering staff behavior recognition method based on machine learning[J]. Computer Systems & Applications, 2019, 28(7): 199-205.
[5]	CHEN Yuqing, XUE Yang. A deep learning approach to human activity recognition based on single accelerometer[C]. 2015 IEEE International Conference on Systems, Man, and Cybernetics, 2015: 1488-1492.
[6]	CHEN Yuwen, ZHONG Kunhua, ZHANG Ju, et al. LSTM networks for mobile human activity recognition[C]. Proceedings of the 2016 International Conference on Artificial Intelligence:Technologies and Applications, 2016: 24-25.
[7]	ORDONEZ F J, ROGGEN D. Deep convolutional and lstm recurrent neural networks for multimodal wearable activity recognition[J]. Sensors, 2016, 16(1): DOI:10.3390/s16010115.
[8]	LI Bing, CUI Wei, WANG Wei, et al. Two-stream convolution augmented transformer for human activity recognition[C]. Proceedings of the AAAI Conference on Artificial Intelligence, 2021, 35(1): 286-293.
[9]	HAMAD R A, KIMURA M, YANG Longzhi, et al. Dilated causal convolution with multi-head self attention for sensor human activity recognition[J]. Neural Computing and Applications, 2021, 33(20): 13 705-13 722.
[10]	韩越林, 王小玉. 多头自注意力在双曲空间下的点击率预测[J]. 北京邮电大学学报, 2021, 44(5):127-132.
	HAN Yuelin, WANG Xiaoyu. Click-through rate prediction of multi-head self-attention in hyperbolic space[J]. Journal of Beijing University of Posts and Telecommunications, 2021, 44(5): 127-132. doi: 10.13190/j.jbupt.2021-017
[11]	KWAPISZ J R, WEISS G M, MOORE S A. Activity recognition using cell phone accelerometers[J]. ACM SigKDD Explorations Newsletter, 2011, 12(2): 74-82. doi: 10.1145/1964897.1964918
[12]	CHATZAKI C, PEDIADITIS M, VAVOULAS G, et al. Human daily activity and fall recognition using a smartphone's acceleration sensor[C]. International Conference on Information and Communication Technologies for Ageing Well and e-Health, 2016: 100-118.
[13]	HASSAN M M, GUMAEI A H, ALOI G, et al. A smartphone-enabled fall detection framework for elderly people in connected home healthcare[J]. IEEE Network, 2019, 33(6): 58-63.

动作	下道	跑	静坐	站立	上道	走
下道	156	1	0	0	16	0
跑	0	715	0	0	0	0
静坐	0	0	140	0	0	0
站立	0	0	0	68	0	0
上道	11	3	1	2	281	2
走	8	0	0	0	4	1 024

动作	下道	跑	静坐	站立	上道	走
下道	136	3	0	0	29	5
跑	0	714	0	0	0	1
静坐	0	0	140	0	0	0
站立	0	0	0	68	0	0
上道	4	3	0	1	281	11
走	2	0	0	0	4	1 030

动作	下道	跑	静坐	站立	上道	走
下道	158	0	0	0	13	2
跑	0	714	0	0	1	0
静坐	0	0	140	0	0	0
站立	0	0	0	68	0	0
上道	11	3	0	1	279	6
走	6	0	0	0	4	1 026

动作	下道	跑	静坐	站立	上道	走
下道	161	3	0	0	8	1
跑	0	715	0	0	0	0
静坐	0	0	140	0	0	0
站立	0	0	0	68	0	0
上道	11	0	2	0	287	0
走	4	0	0	0	1	1 031

铁路工人人体行为识别模型

Human activity recognition model of railway workers

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 8

参考文献 13

相关文章 8

编辑推荐

Metrics

本文评价

[1]	张萌, 韩豫, 刘泽锋. 深度学习下建筑工人高空安全防护装备检测方法[J]. 中国安全科学学报, 2022, 32(5): 140-146.
[2]	郭玮, 袁宏永, 薛明, 魏平岩. SAR影像洪水淹没范围深度学习提取方法[J]. 中国安全科学学报, 2022, 32(4): 177-184.
[3]	段照斌, 杜海龙, 张鹏. 基于QAR2Vec模型的QAR数据特征提取[J]. 中国安全科学学报, 2021, 31(1): 145-152.
[4]	赵欣然, 张琪, 王卫东, 徐志强. 可燃性粉尘云的图像检测方法[J]. 中国安全科学学报, 2020, 30(4): 8-13.
[5]	徐丹, 代勇, 纪军红. 基于卷积神经网络的驾驶人行为识别方法研究[J]. 中国安全科学学报, 2019, 29(10): 12-17.
[6]	汪洋, 王俊刚. 基于深度学习算法的铁路列车运行安全检测^*[J]. 中国安全科学学报, 2018, 28(S2): 41-45.
[7]	佟瑞鹏, 陈策, 崔鹏程, 傅贵, 安宇. 基于深度学习的施工安全泛场景数据获取方法[J]. 中国安全科学学报, 2017, 27(5): 1-6.
[8]	佟瑞鹏, 崔鹏程. 基于深度学习的不安全因素识别和交互分析[J]. 中国安全科学学报, 2017, 27(4): 49-54.