基于多模态的井下登高作业专人扶梯检测方法

孙晴; 杨超宇

doi:10.13272/j.issn.1671-251x.2024010068

基于多模态的井下登高作业专人扶梯检测方法

doi: 10.13272/j.issn.1671-251x.2024010068

孙晴,
杨超宇

安徽理工大学人工智能学院，安徽淮南　232001

基金项目: 国家自然科学基金项目（61873004）。

详细信息

作者简介:
孙晴（2000—），女，河南新乡人，硕士研究生，研究方向为煤矿井下不安全行为智能识别，E-mail：beryl2022@163.com

中图分类号: TD67
计量
- 文章访问数: 98
- HTML全文浏览量: 22
- PDF下载量: 16
- 被引次数: 0
出版历程
- 收稿日期: 2024-01-22
- 修回日期: 2024-05-20
- 网络出版日期: 2024-06-13

A multi-modal detection method for holding ladders in underground climbing operations

School of Artificial Intelligence, Anhui University of Science and Technology, Huainan 232001, China

摘要

摘要: 目前大多数的井下人员不安全行为识别研究侧重于在计算机视觉上提高精度，但井下易出现遮挡、光照不稳定、反光等情况，仅采用计算机视觉技术难以实现对不安全行为的准确识别，尤其登高作业中的爬梯、扶梯等相似动作在识别过程中易被混淆，存在安全隐患。针对上述问题，提出一种基于多模态的井下登高作业专人扶梯检测方法。该方法从视觉和音频2个模态对监控视频数据进行分析。视觉模态方面，采用YOLOv8模型检测登高梯是否存在，如果存在，获得登高梯的位置坐标，并将视频段放入OpenPose算法中进行姿态估计，得到人体的各个骨骼关节点的特征，将这些骨骼关节点序列放入改进的时空注意图卷积网络（SAT−GCN）中，得到人体动作标签及其对应概率。音频模态方面，采用飞桨自动语言识别系统将语音转换为文本，使用双向编码器表示（BERT）模型对文本信息进行特征分析与提取，得到文本标签及其对应的概率。最后将视觉模态与音频模态得到的信息进行决策级融合，判断井下登高作业是否有专人扶梯。实验结果表明：基于骨架数据的动作识别中，优化后的SAT−GCN模型对于扶梯、爬梯、站立3种动作的识别精度分别提升了3.36%，2.83%，10.71%；基于多模态的检测方法比单模态方法具有更高的识别准确率，达到98.29%。
- 登高作业 /
- 专人扶梯 /
- 多模态融合 /
- 视觉模态 /
- 音频模态 /
- YOLOv8 /
- OpenPose /
- SAT−GCN /
- BERT
Abstract: Currently, most research on recognizing unsafe behaviors of underground personnel focuses on improving precision through computer vision. However, underground areas are prone to occlusion, unstable lighting, and reflection, making it difficult to accurately recognize unsafe behaviors using computer vision technology alone. Especially, similar actions such as climbing ladders and holding ladders during climbing operations are easily confused during the recognition process, posing safety hazards. In order to solve the above problems, a multi-modal detection method for holding ladders in underground climbing operations is proposed. This method analyzes surveillance video data from two modalities: visual and audio. In terms of visual modality, the YOLOv8 model is used to detect the presence of ladder. If there is a ladder, the position coordinates of the ladder are obtained, and the video segment is put into the OpenPose algorithm for pose estimation to obtain the features of various skeletal joint points of the human body. These skeletal joint point sequences are then placed into improved spatial attention temporal graph convolutional networks(SAT-GCN) to obtain human action labels and their corresponding probabilities. In terms of audio modality, the PaddlePaddle automatic language recognition system is used to convert speech into text, and the bidirectional encoder representations from transformers (BERT) model is used to analyze and extract the features of text information, so as to obtain the text label and its corresponding probability. Finally, the information obtained from the visual and audio modalities is fused at the decision-making level to determine whether there is a dpersonnel holding ladders for underground climbing operations. The experimental results show that in action recognition based on skeleton data, the optimized SAT-GCN model improves the recognition precision of three types of actions: holding, climbing, and standing by 3.36%, 2.83%, and 10.71%, respectively. The multi-modal detection method has a higher recognition accuracy than the single modal method, reaching 98.29%.
- climbing operation /
- personnel holding ladders /
- multi-modal fusion /
- visual modality /
- audio modality /
- YOLOv8 /
- OpenPose /
- SAT-GCN /
- BERT

HTML全文

图 1 基于多模态的井下登高作业专人扶梯检测方法技术路线

Figure 1. Technical route of multi-modal detection method for holding ladders in underground climbing operations

下载: 全尺寸图片幻灯片

图 2 YOLOv8网络结构

Figure 2. Network architecture of YOLOv8

下载: 全尺寸图片幻灯片

图 3 OpenPose网络结构

Figure 3. Network architecture of OpenPose

下载: 全尺寸图片幻灯片

图 4 井下人员骨骼关节点

Figure 4. Skeletal joint points for underground personnel

下载: 全尺寸图片幻灯片

图 5 ST−GCN结构

Figure 5. Architecture of ST-GCN

下载: 全尺寸图片幻灯片

图 6 SAT−GCN的单个时空图卷积单元组成

Figure 6. Composition of single space-time graph convolutional units in SAT-GCN

下载: 全尺寸图片幻灯片

图 7 SAM结构

Figure 7. Structure of spatial attention module (SAM)

下载: 全尺寸图片幻灯片

图 8 BERT模型结构

Figure 8. Model architecture of BERT

下载: 全尺寸图片幻灯片

图 9 登高作业不安全行为判别流程

Figure 9. Recognition flow of unsafe behavior in climbing operations

下载: 全尺寸图片幻灯片

图 10 可视化训练结果

Figure 10. Visualize training results

下载: 全尺寸图片幻灯片

图 11 不同环境下对登高梯的识别效果

Figure 11. Recognition effects of ladder in different environments

下载: 全尺寸图片幻灯片

图 12 井下登高作业有无专人扶梯行为识别结果

Figure 12. Recognition results of whether there is a personnel holding ladders in underground climbing operations

下载: 全尺寸图片幻灯片

图 13 SAT−GCN模型与ST−GCN模型损失曲线

Figure 13. Loss curves of SAT-GCN model and ST-GCN model

下载: 全尺寸图片幻灯片

表 1 部分文本训练数据

Table 1. Partial text training data

数据	标签
你自己小心点上去就行，不用找人扶梯子，没必要的。我上次就是一个人上去的，没问题的。	0
我来帮你扶一下梯子吧。不用，我熟练得很，没事。	0
我现在要搬个梯子上去维修一下顶板的支架，你来帮我扶着吧。	1
那地面不太平整，梯子放不稳，我来帮你扶着。	1

下载: 导出CSV

表 2 模型对不同动作的识别精度对比

Table 2. Comparison of recognition precision of models on different actions %

动作类别	ST−GCN	SAT−GCN
扶梯	94.31	97.67
爬梯	72.89	75.72
站立	75.00	85.71

下载: 导出CSV

表 3 模型在自建数据集上的实验对比

Table 3. Experimental comparison of models on self-builting datasets %

模型	准确率
ST−GCN	80.39
SAT−GCN	82.35
VA−RNN	77.25
2s−AGCN	84.97
基于多模态融合的井下登高作业专人扶梯检测模型	98.29

下载: 导出CSV

参考文献(22)

[1]	张瑜,冯仕民,杨赛烽,等. 矿工不安全行为影响因素本体构建与推理研究[J]. 煤矿安全,2019,50(5):300-304. ZHANG Yu,FENG Shimin,YANG Saifeng,et al. Ontology construction and reasoning research on influencing factors of miners' unsafe behavior[J]. Safety in Coal Mines,2019,50(5):300-304.
[2]	登高作业操作规程[EB/OL]. (2021-11-08)[2023-10-08]. https://www.mkaq.org/html/2021/11/08/593666.shtml. Operation procedures for climbingoperations[EB/OL]. (2021-11-08)[2023-10-08]. https://www.mkaq.org/html/2021/11/08/593666.shtml.
[3]	刘浩,刘海滨,孙宇,等. 煤矿井下员工不安全行为智能识别系统[J]. 煤炭学报,2021,46(增刊2):1159-1169. LIU Hao,LIU Haibin,SUN Yu,et al. Intelligent recognition system of unsafe behavior of underground coal miners[J]. Journal of China Coal Society,2021,46(S2):1159-1169.
[4]	饶天荣,潘涛,徐会军. 基于交叉注意力机制的煤矿井下不安全行为识别[J]. 工矿自动化,2022,48(10):48-54. RAO Tianrong,PAN Tao,XU Huijun. Unsafe action recognition in underground coal mine based on cross-attention mechanism[J]. Journal of Mine Automation,2022,48(10):48-54.
[5]	王宇,于春华,陈晓青,等. 基于多模态特征融合的井下人员不安全行为识别[J]. 工矿自动化,2023,49(11):138-144. WANG Yu,YU Chunhua,CHEN Xiaoqing,et al. Recognition of unsafe behaviors of underground personnel based on multi modal feature fusion[J]. Journal of Mine Automation,2023,49(11):138-144.
[6]	赵登阁,智敏. 用于人体动作识别的多尺度时空图卷积算法[J]. 计算机科学与探索,2023,17(3):719-732. doi: 10.3778/j.issn.1673-9418.2106102 ZHAO Dengge,ZHI Min. Spatial multiple-temporal graph convolutional neural network for human action recognition[J]. Journal of Frontiers of Computer Science and Technology,2023,17(3):719-732. doi: 10.3778/j.issn.1673-9418.2106102
[7]	LI Peilin,WU Fan,XUE Shuhua,et al. Study on the interaction behaviors identification of construction workers based on ST-GCN and YOLO[J]. Sensors,2023,23(14). DOI: 10.3390/S23146318.
[8]	SHI Xiaonan,HUANG Jian,HUANG Bo. An underground abnormal behavior recognition method based on an optimized Alphapose-ST-GCN[J]. Journal of Circuits,Systems and Computers,2022,31(12). DOI: 10.1142/S0218126622502140.
[9]	苏晨阳,武文红,牛恒茂,等. 深度学习的工人多种不安全行为识别方法综述[J]. 计算机工程与应用,2024,60(5):30-46. SU Chenyang,WU Wenhong,NIU Hengmao,et al. Review of deep learning approaches for recognizing multiple unsafe behaviors in workers[J]. Computer Engineering and Applications,2024,60(5):30-46.
[10]	YAN Sijie,XIONG Yuanjun,LIN Dahua. Spatial temporal graph convolutional networks for skeleton-based action recognition[C]. AAAI Conference on Artificial Intelligence,New Orleans,2018:7444-7452.
[11]	SONG Sijie,LAN Cuiling,XING Junliang,et al. Spatio-temporal attention-based LSTM networks for 3D action recognition and detection[J]. IEEE Transactions on Image Processing,2018,27(7):3459-3471. doi: 10.1109/TIP.2018.2818328
[12]	CAO Zhe,SIMON T,WEI S E,et al. Realtime multi-person 2D pose estimation using part affinity fields[C]. IEEE Conference on Computer Vision and Pattern Recognition,Honolulu,2017:1302-1310.
[13]	许奇珮. 基于ST−GCN的人体骨架动作识别方法研究[D]. 长春:长春工业大学,2023. XU Qipei. Research on human skeleton action recognition method based on ST-GCN[D]. Changchun:Changchun University of Technology,2023.
[14]	DEVLIN J,CHANG Mingwei,LEE K,et al. BERT:pre-training of deep bidirectional transformers for language understanding[EB/OL]. (2018-10-11)[2023-10-08]. https://doi.org/10.48550/arXiv.1810.04805.
[15]	LIU Shu,QI Lu,QIN Haifeng,et al. Path aggregation network for instance segmentation[C]. IEEE Conference on Computer Vision and Pattern Recognition,Salt Lake City,2018:8759-8768.
[16]	李雯静,刘鑫. 基于深度学习的井下人员不安全行为识别与预警系统研究[J]. 金属矿山,2023(3):177-184. LI Wenjing,LIU Xin. Research on underground personnel unsafe behavior identification and early warning system based on deep learning[J]. Metal Mine,2023(3):177-184.
[17]	刘耀,焦双健. ST−GCN在建筑工人不安全动作识别中的应用[J]. 中国安全科学学报,2022,32(4):30-35. LIU Yao,JIAO Shuangjian. Application of ST-GCN in unsafe action identification of construction workers[J]. China Safety Science Journal,2022,32(4):30-35.
[18]	WOO S,PARK J,LEE J Y,et al. CBAM:convolutional block attention module[C]. European Conference on Computer Vision,Cham,2018:3-19.
[19]	景永霞,苟和平,刘强. 基于BERT语义分析的短文本分类研究[J]. 兰州文理学院学报(自然科学版),2023,37(6):46-49. JING Yongxia,GOU Heping,LIU Qiang. Classification study on online short text based on BERT semantic analysis[J]. Journal of Lanzhou University of Arts and Science(Natural Sciences),2023,37(6):46-49.
[20]	姜长三,曾桢,万静. 多源信息融合研究进展综述[J]. 现代计算机,2023,29(18):1-9,29. JIANG Changsan,ZENG Zhen,WAN Jing. A review of research advances in multi-source information fusion[J]. Modern Computer,2023,29(18):1-9,29.
[21]	ZHANG Pengfei,LAN Cuiling,XING Junliang,et al. View adaptive neural networks for high performance skeleton-based human action recognition[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,2019,41(8):1963-1978. doi: 10.1109/TPAMI.2019.2896631
[22]	SHI Lei,ZHANG Yifan,CHENG Jian,et al. Two-stream adaptive graph convolutional networks for skeleton-based action recognition[C]. IEEE/CVF Conference on Computer Vision and Pattern Recognition,Long Beach,2019:12018-12027.