1 School of Automation and Information Engineering,Sichuan University of Science & Engineering,Zigong 643000,China
2 Artificial Intelligence Key Laboratory of Sichuan Province,Zigong 643000,China
Show less
文章历史+
Received
Published
Revised
2025-02-21
2026-02-08
2025-10-25
Online Date
2026-09-21
PDF (20013K)
摘要
堆煤是输送机常见故障之一,为了保障煤矿工业生产的安全,需要对煤矿井下输送机的堆煤情况进行检测。然而现有的检测方法存在容易误触、检测可靠性较差等缺点,针对这些问题提出一种基于Transformer统一多尺度时序卷积(unified multi-scale temporal ConvTransformer,UnMS-TCT)网络用于输送机堆煤检测。首先融合RGB帧和光流帧提取的特征,使网络更全面地建模时空关系;然后在时序编码器中,将动态位置嵌入(dynamic position embedding,DPE),多头关系聚合器(multi-head relation aggregator,MHRA)以及多层感知机(multilayer perceptron,MLP)组成的全局模块,交叉注意力(cross-attention,CA)组成的局部模块,以交替方式形成全局-局部关系模块,增强多尺度下获取全局和局部时间关系的能力;其次利用残差全局-局部融合(residual global and local fusion,ResGLFus)模块融合多尺度特征,有效地提高融合过程的稳定性,最终实现高精度堆煤预测。实验结果表明:该方法能够实现对输送机堆煤的检测,mAP达到98.17%。
Abstract
Coal pile is one of the common faults of conveyors. In order to ensure the safety of coal mine industrial production,it is necessary to detect the coal pile situation of underground conveyors in coal mines. However,the existing detection methods have the disadvantages of easy false touch and poor detection reliability. To solve these problems,a unified multi-scale temporal ConvTransformer (UnMS-TCT) network based on Transformer was proposed for conveyor coal pile detection. Firstly,the features extracted from RGB frames and optical flow frames were fused to enable the network to model the spatiotemporal relationship more comprehensively. Then,in the temporal encoder,the global module composed of dynamic position embedding (DPE),multi-head relation aggregator (MHRA) and multilayer perceptron (MLP) and the local module composed of cross-attention (CA) were alternately formed into a global-local relationship module to enhance the ability to obtain global and local temporal relationships at multiple scales. Secondly,the residual global-local fusion (ResGLFus) module was used to fuse multi-scale features,effectively improving the stability of the fusion process,and finally achieving high-precision coal pile prediction. The experimental results show that this method can detect coal piles on conveyors,and the mAP reaches 98.17%.
ZHANG Jin-yang, WU Hao, WANG Ming-yao, YANG Fu-jie.
Coal Pile Detection in Underground Conveyor of Coal Mine Based on Unified Multi-scale Temporal ConvTransformer Network[J].
Science Technology and Engineering, 2026, 26(4): 1415-1422 DOI:10.12404/j.issn.1671-1815.2501099
YuanLiang. Theory and technology considerations on high-quality development of coal main energy security in China[J]. Bulletin of Chinese Academy of Sciences, 2023, 38(1): 11-22.
MiaoChangyun, LiJia. Machine vision-based coal pile detection at drop port of belt conveyor[J]. Journal of Liaoning Technical University(Natural Science), 2023, 42(5): 617-624.
HuoJianjun. Mechanical structure design of anti-stacking coal device for mining belt conveyor[J]. Journal of Mine Automation, 2024, 50(S2): 183-185, 194.
ZhangHeping. Research on fault monitoring system of mine belt conveyor based on machine vision[J]. Coal Science & Technology Magazine, 2020, 41(1): 89-92.
YouLei, ZhuXinglin, QinWei, et al. Coal stacking identification method of belt conveyor based on surface reconstruction[J]. Journal of Mine Automation, 2021, 47(6): 45-50.
[17]
HenriqueD F, FernandoL, RamonR C, et al. ROSI: a mobile robot for inspection of belt conveyor[J]. IFAC Papers Online, 2020, 53(2): 10031-10036.
FangXinqiu, WuYang, SongYang, et al. Research on fault monitoring of belt conveyor based on FBG sensor[J]. Coal Science and Technology, 2025, 53(1): 326-340.
MaoQinghua, GuoWenjin, ZhaiJiao, et al. Research on video AI recognition technology for abnormal state of coal mine belt conveyors[J]. Journal of Mine Automation, 2023, 49(9): 36-46.
TangJun, LiJingzhao, ShiQing, et al. Video real-time detection of bulk material accumulation on belt conveyor[J]. Journal of Mine Automation, 2022, 48(10): 62-68, 75.
XuShichang, ChengGang, YuanDunpeng, et al. Belt conveyor deviation and coal stacking monitoring method based on three dimensional point cloud[J]. Journal of Mine Automation, 2022, 48(9): 8-15, 24.
TuDuorui, LiJingzhao, ChenRuiyun, et al. Coal stacking detection system of belt conveyor based on multi-information fusion[J]. Auto-mation Application, 2024, 65(17): 152-154, 158.
LiDong. Research on monitoring technology belt conveyor runout and coal stacking based on 3D point cloud[J]. Mechanical Manage-ment and Development, 2024, 39(11): 291-293.
WangZhenhua, DuanZiqiang, ZhangRuifeng, et al. Application of the intelligent coal flow monitoring system in the main transportation line of Buliangou coal mine[J]. Journal of Intelligent Mine, 2024, 5(10): 60-66.
[32]
VaswaniA. Attention is all you need[J]. Advances in Neural Information Processing Systems, 2017, 28(3): 14.
ChenYu, LiuHui, LiangDongsheng, et al. Person reidentification based on pose estimation and Transformer model[J]. Science Technology and Engineering, 2024, 24(12): 5051-5058.
GaoJia, CaiWenhao, ZhaoJunli, et al. ViTAU: facial paralysis recognition and analysisbased on vision transformer and facialaction units[J]. Chinese Journal of Engineering, 2025, 47(2): 351-363.
[37]
JiaoJ, TangY M, LinK Y, et al. Dilateformer: multi-scale dilated transformer for visual recognition[J]. IEEE Transactions on Multimedia, 2023, 25: 8906-8919.
ZhaoLiang, ZhouJikai. Vision Transforme-r based on reconfigurable Gaussian self-attention[J]. Acta Automatica Sinica, 2023, 49(9): 1976-1988.
[40]
SunW, MaY, WangR. KNN attention-basedvideo vision transformer for action recognition[J]. Neurocomputing, 2024, 574: 127256.
[41]
DaiY M. Attentional feature fusion[C]// Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision. New York: IEEE, 2021: 1-5.
[42]
LiuT, MaY, YangW, et al. Spatial-temporal interaction learning based two-stream network for action recognition[J]. Information Sciences, 2022, 606: 864-876.
[43]
LiK, WangY, GaoP, et al. Uniformer: unified transformer for efficient spatiotemporal representation learning[J]. arXiv Preprint, arXiv: 2201.04676, 2022.
[44]
ChenC F, FanQ, PandaR. Crossvit: Cross-attention multi-scale vision transformer for image classification[C]// Proceedings of the IEEE/CVF International Conference on Computer Vision. New York: IEEE, 2021: 357-366.
[45]
DaiR, DasS, MinciulloL, et al. PDAN: pyramid dilated attention network for action detection[C]// Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision. New York: IEEE, 2021: 2970-2979.
[46]
PiergiovanniA J, RyooM. Temporal Gaussian mixture layer for videos[C]// International Conference on Machine learning. Huston: PMLR, 2019: 5152-5161.
[47]
PiergiovanniA J, RyooM S. Learning latent super-events to detect multiple activities in videos[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2018: 5304-5313.
[48]
LiuX, WangQ, HuY, et al. End-to-end temporal action detection with transformer[J]. IEEE Transactions on Image Processing, 2022, 31: 5427-5441.
[49]
TanJ, TangJ, WangL, et al. Relaxed transformer decoders for direct action proposal generation[C]// Proceedings of the IEEE/CVF International Conference on Computer Vision. New York: IEEE, 2021: 13526-13535.