基于计算机视觉的高处临边作业安全巡检

doi:10.16265/j.cnki.issn1003-3033.2023.09.0202

中国安全科学学报 ›› 2023, Vol. 33 ›› Issue (9): 69-75.doi: 10.16265/j.cnki.issn1003-3033.2023.09.0202

基于计算机视觉的高处临边作业安全巡检

李华(), 吴立舟^**(), 薛曦澄, 钟兴润

西安建筑科技大学资源工程学院,陕西西安 710055

收稿日期:2023-03-17 修回日期:2023-06-19 出版日期:2023-09-28
通讯作者:
**吴立舟(2000—),男,陕西汉中人,硕士研究生,主要研究方向为智能施工与安全管理。E-mail:352661069@qq.com。
作者简介:
李华 (1979—),女,陕西西安人,博士,副教授,硕士生导师,主要从事企业风险评估与安全管理、建筑安全监测与监控、公共安全与应急管理方面的研究。E-mail:lihua@xauat.edu.cn。
钟兴润讲师
基金资助:
陕西省建设厅科技发展计划项目(2020-K32); 西安建科大工程技术项目(XAJD-YF23N010)

Computer vision based safety inspection of high abutting edges

LI Hua(), WU Lizhou^**(), XUE Xicheng, ZHONG Xingrun

School of Resources Engineering,Xi'an University of Architecture and Technology, Xi'an Shaanxi 710055,China

Received:2023-03-17 Revised:2023-06-19 Published:2023-09-28

摘要/Abstract

摘要：

为解决建筑施工现场高处作业中安全管理人员巡检工作时存在的危险性高、效率低、识别场景复杂等问题,利用卷积神经网络(CNN)技术,提出一种高处临边安全防护装备的巡检方法,该方法结合计算机视觉,检测高处临边人员安全帽、安全带等安全防护装备的佩戴情况以及防护网是否破损;同时在YOLOv5算法基础上修改注意力模型,并开发轻量化检测软件。结果表明:轻量化后,模型尺寸降低到1.9 MB,相较于修改前减小86.8%。在图形处理器(GPU)运行环境下单帧图片检测时间优化到40 ~50 ms,相较于修改前减少65%~80%,大幅提高检测速度。

关键词: 计算机视觉, 高处临边, 安全巡检, 轻量化, 施工现场, YOLOv5

Abstract:

To address the issues of high danger, low efficiency, and complex recognition scenarios in the safety inspection work of construction site managers, a new detection method for high-altitude safety protection equipment was proposed using convolutional neural network (CNN) technology. This method combines computer vision to detect the wearing status of personal protective equipment such as safety helmets and harnesses, as well as any damage to protective nets. Additionally, based on the YOLOv5 algorithm, the attention model was modified and a lightweight detection software was developed. The results indicate that after lightweighting, the model size decreased to 1.9 MB, a reduction of 86.8% compared to before the modification. Under graphics processing unit(GPU) operating conditions, the single-frame image detection time was optimized to 40-50 ms, representing a reduction of 65%-80% compared to before, greatly improving the detection speed.

Key words: computer vision, elevated proximity, safety inspection, lightweight, construction site, YOLOv5

李华, 吴立舟, 薛曦澄, 钟兴润. 基于计算机视觉的高处临边作业安全巡检[J]. 中国安全科学学报, 2023, 33(9): 69-75.

LI Hua, WU Lizhou, XUE Xicheng, ZHONG Xingrun. Computer vision based safety inspection of high abutting edges[J]. China Safety Science Journal, 2023, 33(9): 69-75.

图/表 9

图1

图2

表1

表2

图3

表3

图4

图5

图6

参考文献 13

[1]	张鸿辉, 李润求. 2012—2018年建筑施工事故统计分析及对策[J]. 科技创新与应用, 2020(31):135-137, 139.
	ZHANG Honghui, LI Runqiu. Statistical analysis and countermeasures of construction accidents in years 2012-2018[J]. Technology Innovation and Application, 2020(31):135-137, 139.
[2]	XU Maozeng, NIE Xiuying, LI Heng, et al. Smart construction sites: a promising approach to improving on-site HSE management performance[J]. Journal of Building Engineering, 2022, 49: DOI: 10.1016/j.jobe.2022.104007.
[3]	郁润, 张知田, 郭红领. 数据驱动的施工现场工人不安全行为管理方法[J]. 中国安全科学学报, 2019, 29(8):130-135. doi: 10.16265/j.cnki.issn1003-3033.2019.08.021
	YU Run, ZHANG Zhitian, GUO Hongling. A data-driven management approach of construction workers' unsafe behaviors[J]. China Safety Science Journal, 2019, 29(8):130-135. doi: 10.16265/j.cnki.issn1003-3033.2019.08.021
[4]	冯领香, 林新超, 顾亦凡, 等. 建筑施工安全绩效与安全投入关系的ABM模拟[J]. 中国安全科学学报, 2017, 27(11):163-168. doi: 10.16265/j.cnki.issn1003-3033.2017.11.028
	FENG Lingxiang, LIN Xinchao, GU Yifan, et al. Studying relationship between safety inputs and safety performance of construction by agent-based model[J]. China Safety Science Journal, 2017, 27(11):163-168. doi: 10.16265/j.cnki.issn1003-3033.2017.11.028
[5]	裴慧坤, 杨兴, 陈城, 等. 架空输电线路无人机巡检视频中避雷器计数器目标检测[J]. 武汉大学学报:工学版, 2022, 55(10):1072-1080.
	PEI Huikun, YANG Xing, CHEN Cheng, et al. Lighting arrester counter target detection for UAV inspection videoof overhead transmission lines[J]. Engineering Journal of Wuhan University, 2022, 55(10):1072-1080.
[6]	黄建城, 徐昆, 董湛波. 智慧工地管理平台系统架构与实现[J]. 建筑经济, 2021, 42(11):25-30.
	HUANG Jiancheng, XU Kun, DONG Zhanbo. Research and implementation of the system architecture of smart construction site management platform[J]. Construction Economy, 2021, 42(11):25-30.
[7]	JAYARAJU P, SOMASUNDARAM K, SUPRAKASH A S, et al. A deep learning-image based approach for detecting cracks in buildings[J]. Traitement du Signal, 2022, 39(4) :1429-1434. doi: 10.18280/ts
[8]	金沙沙, 龙伟, 胡灵犀, 等. 多目标检测与跟踪算法在智能交通监控系统中的研究进展[J]. 控制与决策, 2023, 38(4):890-901.
	JIN Shasha, LONG Wei, HU Lingxi, et al. Research progress of detection and multi-object tracking algorithm inintelligent traffc monitoring system[J]. Control and Decision, 2023, 38(4):890-901.
[9]	WANG Jingjing, LUAN Zhenye, YU Zishu, et al. Superpixel segmentation with squeeze-and-excitation networks[J]. Signal, Image and Video Processing, 2022, 16(5) :1161-1168. doi: 10.1007/s11760-021-02066-2
[10]	WANG Qilong, WU Banggu, ZHU Pengfei, et al. ECA-Net: efficient channel attention for deep convolutional neural networks[C]. Proceedings of the Ieee Computer Society Conference on Computer Vision and Pattern Recognition, 2020:11 531-11 539.
[11]	XIE Chao, ZHU Hongyu, FEI Yeqi. Deep coordinate attention network for single image super‐resolution[J]. IET Image Processing, 2021, 16(1):273-284. doi: 10.1049/ipr2.v16.1
[12]	JIN Han, YANG Yonghao. L-Net: lightweight and fast object detector-based ShuffleNetV2[J]. Journal of Real-time Image Processing, 2021, 18(6):2527-2538. doi: 10.1007/s11554-021-01145-4
[13]	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. doi: 10.16265/j.cnki.issn1003-3033.2022.05.1141
	ZHANG Meng, HAN Yu, LIU Zefeng. Detection method of high-altitude safety protective equipment for construction workers based on deep learning[J]. China Safety Science Journal, 2022, 32(5):140-146. doi: 10.16265/j.cnki.issn1003-3033.2022.05.1141

类别	解释
TP(True Positives)	实际为正例且被分类器划分为正例的实例数
FP(False Positives)	实际为负例但被分类器划分为正例的实例数
FN(False Negatives)	实际为正例但被分类器划分为负例的实例数
TN(True Negatives)	实际为负例且被分类器划分为负例的实例数

训练方法	P	R	mAP@ 0.5	mAP@ 0.5:0.95	模型内存/ MB	运算时间/ h
YOLOv5	0.77	0.73	0.74	0.41	14.4	6.89
YOLOv5-SE	0.82	0.70	0.72	0.40	15.4	7.00
YOLOv5-CBAM	0.83	0.68	0.73	0.40	14.7	6.96
YOLOv5-ECA	0.84	0.70	0.73	0.40	14.7	7.04
YOLOv5-CA	0.83	0.68	0.71	0.40	14.4	7.23

训练方法	P	模型内存/MB	运算时间/h
YOLOv5	0.77	14.4	6.886
YOLOv5-ECA	0.84	14.7	7.038
YOLOv5-ShuffleNerV2	0.83	1.9	6.435

基于计算机视觉的高处临边作业安全巡检

Computer vision based safety inspection of high abutting edges

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 9

参考文献 13

相关文章 8

编辑推荐

Metrics

本文评价

[1]	王瑶涵, 宋泽阳, 张利冬. 基于卷积神经网络的安全标识分类算法研究[J]. 中国安全科学学报, 2023, 33(S1): 263-269.
[2]	张磊, 李熙尉, 燕倩如, 王浩盛, 雷伟强. 基于改进YOLOv5s的综采工作面人员检测算法[J]. 中国安全科学学报, 2023, 33(7): 82-89.
[3]	王喆, 李享, 杨栋梁, 刘丹. 基于YOLOv5s的城镇森林交界域火灾探测模型[J]. 中国安全科学学报, 2023, 33(6): 152-158.
[4]	李继超, 郭聖煜, 孔刘林, 原毅璨, 孙阳. 施工现场火焰检测和预警机器人设计及应用[J]. 中国安全科学学报, 2021, 31(4): 141-146.
[5]	李伟, 阳富强. 基于SLP的地铁施工场地安全布局优化方案[J]. 中国安全科学学报, 2019, 29(1): 161-166.
[6]	张海涛，吴伟巍，李响，李启明. 基于循证方法的施工现场安全事故预防体系研究[J]. 中国安全科学学报, 2012, 22(12): 17-.
[7]	徐海祥，吴卫国，余晋刚，田金文，郑元洲. 船-桥避碰监测预警系统研究[J]. 中国安全科学学报, 2011, 21(7): 151-.
[8]	贺耀荣，汪莉，李祥，宋存义. 基于BP网络的建筑安装施工现场安全综合评价的研究[J]. 中国安全科学学报, 2004, 14(12): 100-.