Railway fastener inspection faces challenges in both detection speed and accuracy due to inconspicuous defect features and complex background contrast. To achieve faster and more precise fastener identification, an improved YOLOv10-based detection algorithm was proposed. Specifically, an attention mechanism (SPPF-Attention) was introduced into the SPPF module to enhance feature recalibration and improve focus on critical regions. Concurrently, a dual attention block (C2f-DAB) was embedded in the C2f module, which combines channel-space attention with parallel attention to strengthen multi-dimensional feature extraction. Experimental results on a self-built railway fastener dataset demonstrates that the original YOLOv10 achieves an average precision of 88.1%. The SPPF-Attention variant achieves an improvement to 89.3%, while the C2f-DAB implementation gains 2.1% over the baseline. All three models maintained real-time performance (42, 41, and 39 frames/s, respectively), meeting practical inspection requirements. The effectiveness of attention mechanism and dual attention block network model in railway fastener detection task was verified through comparative experiments, providing theoretical basis and practical reference for research in related fields.
MENG Jian-jun, YAO Li-qing, LÜ De-fang, SHI Rui, LI Yi-xuan.
Railway Fastener Detection Algorithm Based on SPPF-Attention and C2f-DAB[J].
Science Technology and Engineering, 2026, 26(9): 3859-3866 DOI:10.12404/j.issn.1671-1815.2502578
NiuDao’an, ChenDongsheng, ZhaoGang, et al. Innovative research on railway infrastructure inspection and monitoring system for modernization construction[J]. China Railway, 2024(10): 1-9.
[3]
ShahM S. 基于改进的YOLO架构、Inception-Resnet-V2和XGazeNet的司机(面部、头部姿态和眼球注视)监控系统[D]. 西安: 长安大学, 2024.
[4]
ShahM S. Driver(face, head pose, and eye gaze) monitoring system based on improved YOLO architecture, Inception-ResNet-V2, and XGazeNet[D]. Xi’an: Chang’an University, 2024.
LiuChunjuan, XinYuqiang, WuXiaosuo, et al. Dual attention-guided U-Net++ for semantic segmentation of remote sensing images[J/OL]. Journal of Beijing University of Aeronautics and Astronautics, 1-13 [2025-03-02]. DOI: 10.13700/j.bh.1001-5965.2024.0122.
ChenQihuai, LinTianliang, MaRonghua, et al. Autonomous Driving Decision Network for Crawler Excavators Based on Multimodal Learning[J]. Journal of Mechanical Engineering, 2025, 61(4): 380-388.
LiuTan, LiuNa, LiuGuiping, et al. Applications and research directions of deep learning in wind power forecasting: a review[J]. Journal of Frontiers of Computer Science and Technology, 2025, 19(3): 602-622.
[11]
龙炎. 基于深度学习的铁路扣件检测系统的研究[D]. 北京: 北京交通大学, 2018.
[12]
LongYan. Research on railway fastener detection system based on deep learning[D]. Beijing: Beijing Jiaotong University, 2018.
ChenZihao. Research on defect identification method for heavy-haul railway fasteners based on deep learning[D]. Beijing: Beijing Jiaotong University, 2023.
WeiRuoyu, LiShuting, WuSongrong, et al. Defect detection of railway fasteners based on improved YOLOv3 algorithm[J]. Railway Standard Design, 2020, 64(12): 30-36.
GaoJialin, BaiTangbo, YaoDechen, et al. Railway fastener detection based on improved YOLOv4 algorithm[J]. Science Technology and Engineering, 2022, 22(7): 2872-2877.
[19]
HuJ, QiaoP, LüH, et al. High speed railway fastener defect detection by using improved YoLoX-Nano model[J]. Sensors, 2022, 22: 8399-8415.
ZhangYuan, MengJianjun, LüDefang, et al. Rail fastener defect detection system based on lightweight model[J]. Instrument Technique and Sensor, 2023(11): 96-101, 106.
[24]
WangS H, HaoX L. YOLO-SK: a lightweight multiscale object detection algorithm[J]. Heliyon, 2024, 10(2): e24143.
WuSongying, LiuLinya, WanQianghua, et al. Railway fastener looseness detection algorithm based on wavelet packet energy spectrum and improved bp neural network[J]. Journal of East China Jiaotong University, 2023, 40(6): 69-78.
[27]
WuL B, GuY H, WuW H, et al. Remote sensing rotating object detection based on multi scale feature extraction[J]. Laser & Optoelectronics Progress, 2023, 60(12): 1228010.
[28]
NguyenT P, NguyenL G, BuiD D. LW-UAV-YOLOv10: a lightweight model for small UAV detection on infrared data based on YOLOv10[J]. Geomatica, 2025, 77(1): 100049.
[29]
WangA, ChenH, LiuL, et al.YOLOv10: real-time end-to-end object detection[C]// Proceedings of the 38th International Conference on Neural Information Processing Systems. Vancouver, BC: ACM, 2024: 107984-108011.
[30]
YeerjiangA, WangZ, HuangX, et al. YOLOv1 to YOLOv10: a comprehensive review of YOLO variants and their application in medical image detection[J]. Journal of Artificial Intelligence Practice, 2024(3). DOI: 10.23977/JAIP.2024.070314.
DuHong, GuChenyu, ZhangXiaozheng, et al. Vehicle detection in complex weather conditions based on YOLOv10-vehicle algorithm[J]. Journal of Jilin University (Engineering and Technology Edition), 2025, 55(10): 3309-3318.
[33]
LuS Y, LiuM Z, YinL R, et al. The multi-modal fusion in visual question answering: a review of attention mechanisms[J]. PeerJ Computer Science, 2023, 9: e1400.
LiWeigang, ShaoJiale, TianZhiqiang. Point cloud classification and segmentation network based on dual attention mechanism and multi-scale fusion[J]. Journal of Computer Applications, 2025, 45(9): 3003-3010.
ZhangXiu, JiKe, MaKun, et al. Event matching algorithm based on multi-dimensional event feature fusion and semantic feature interaction[J]. Data Analysis and Knowledge Discovery, 2025, 9(4): 111-122.