Restricted accessReview articleFirst published online 2025
Automatic detection and localization of surface defects in high-speed railway ballastless track based on cascaded group attention and optoelectronic encoder
Surface defects in ballastless tracks pose potential risks to the safety and stability of high-speed railway operations. However, the complex detection environment at night often leads to low computational efficiency and detection accuracy in structural damage assessment. To address these issues, this article proposes a lightweight detection and localization method for surface damage in ballastless tracks, deployed on an intelligent track inspection vehicle. The method enhances the network model’s ability to extract shallow and deep damage features under nighttime conditions, while mitigating the harmful gradient effects caused by low-quality samples. Besides, the method uses photoelectric encoding technology to achieve precise mileage localization of surface defects. Compared to the You Only Look Once (YOLO)v7 model, the YOLO-Track model achieves a precision of 98.2%, reflecting a 7.68% improvement, and reduces the parameter size by 33.80%, down to 23.05M. The YOLO-Track model with a dynamic nonmonotonic mechanism loss function achieves a mean average precision @0.75 of 81.4%, surpassing the traditional complete intersection over union loss function model by 13.06%. Finally, the effectiveness of the proposed method is validated through field tests.
LiXRenJLiuX, et al. Multi-field, time-varying behavior and cracking mechanisms of early-age concrete in balastless track beds. J Zhejiang Univ Sci A2024; 6(3): 212–225.
2.
LuC.A discussion on technologies for improving the operational speed of high-speed railway networks. Transp Saf Environ2019; 1(1): 22–36.
3.
LuCRenZMaC.Study on the technologies development trend of high speed EMUs. High Speed Rail2023; 1(1): 1–5.
4.
YeWRenJZhangP, et al. Review of integrated full life cycle data management and application of the slab tracks. Intell Transp Infrastruct2022; 1: 1–14.
5.
JuanjuanRWengaoLJialeL, et al. Performance deterioration and structural state diagnosis of slab tracks for high-speed railways: a review. Eng Fail Anal2024; 158: 107955.
6.
ZhanYDaiXYangE, et al. Convolutional neural network for detecting railway fastener defects using a developed 3D laser system. Int J Rail Transp2021; 9(5): 424–444.
7.
FengQSunKChenH-P, et al. Long-term prediction of fatigue crack growth in ballastless track of high-speed railway due to cyclic train load. Constr Build Mater2021; 292: 123375.
8.
YeWRenJZhangAA, et al. Automatic pixel-level crack detection with multi-scale feature fusion for slab tracks. Comput Aided Civil Infrastruct Eng2023; 38(16): 1–18.
9.
CaiXTangXPanS, et al. Intelligent recognition of defects in high-speed railway slab track with limited dataset. Comput Aided Civil Infrastruct Eng2024; 39(6): 911–928.
10.
ChellaswamyCKrishnasamyMBalajiL, et al. Optimized railway track health monitoring system based on dynamic differential evolution algorithm. Measurement2020; 152: 1–18.
11.
DuWRenJZhangK, et al. Two-stage identification of interlayer contact loss for CRTS III prefabricated slab track based on multi-index fusion. J Zhejiang Univ Sci A2023; 24(6): 497–515.
12.
BaiTYangJXuG, et al. An optimized railway fastener detection method based on modified Faster R-CNN. Measurement2021; 182: 1–9.
13.
HuJQiaoPLvH, et al. High speed railway fastener defect detection by using improved YoLoX-nano model. Sensors2022; 22(21): 8399.
14.
ZhaiGXuYSpencerBF.Bidirectional graphics-based digital twin framework for quantifying seismic damage of structures using deep learning networks. Struct Health Monit2024; 24(1): 86–110.
15.
JiZZhaoJLiuJ, et al. ELCT-YOLO: an efficient one-stage model for automatic lung tumor detection based on CT images. Mathematics2023; 11(10): 2344.
16.
JiZWuYZengX, et al. Lung nodule detection in medical images based on improved YOLOv5s. IEEE Access2023; 11: 76371–76387.
17.
YeWRenJLuC, et al. Intelligent detection of fastener defects in ballastless tracks based on deep learning. Autom Constr2024; 159: 105280.
18.
WuYMengFQinY, et al. UAV imagery based potential safety hazard evaluation for high-speed railroad using Real-time instance segmentation. Adv Eng Inform2023; 55: 101819.
19.
GirshickRDonahueJDarrellT, et al. Rich feature hierarchies for accurate object detection and semantic segmentation. In: Proceedings of the IEEE conference on computer vision and pattern recognition, Columbus, OH, USA, 2014, pp.580–587.
20.
GirshickR. Fast R-CNN: fast region-based convolutional networks for object detection. In: International conference on computer vision, ICCV, Santiago, Chile, 7–13 December 2015, pp.1–6.
21.
RenSHeKGirshickR, et al. Faster R-CNN: towards real-time object detection with region proposal networks. Adv Neural Inform Process Syst2015; 28: 1–9.
22.
LiuWAnguelovDErhanD, et al. SSD: single shot multibox detector. In: Computer vision–ECCV 2016: 14th European conference, Amsterdam, The Netherlands, October 11–14, 2016, Proceedings, Part I 14, Springer, 2016, pp.21–37.
23.
RedmonJDivvalaSGirshickR, et al. You only look once: unified, real-time object detection. In: Proceedings of the IEEE conference on computer vision and pattern recognition, Las Vegas, Nevada, 26 June–1 July 2016, pp.779–788.
24.
WangCYBochkovskiyALiaoHYM. YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors. arXiv Preprint2022; 2022: arXiv:2207.02696.
25.
PriyaDKRamkumarMMenakaD. Rail fastener fault detection based on enhanced YOLOv7 model. In: 2023 International conference on circuit power and computing technologies (ICCPCT). IEEE, Kollam, India, 10–11 August 2023, pp.1681–1689.
26.
BaiTLvBWangY, et al. Crack detection of track slab based on RSG-YOLO. IEEE Access2023; 11: 124004–124013.
27.
ChuZZhangJWangC, et al. YOLO-O2E: a variant YOLO model for anomalous rail fastening detection. Comput Mater Con2024; 80(1): 1143–1161.
28.
DangCWangZHeY, et al. The accelerated inference of a novel optimized yolov5-lite on low-power devices for railway track damage detection. IEEE Access2023; 11: 134846–134865.
29.
GuoFQianYLiuJ, et al. Pavement crack detection based on transformer network. Autom Constr2023; 145: 104646.
30.
VaswaniAShazeerNParmarN, et al. Attention is all you need. Adv Neural Inf Process Syst2017; 30: 1–6.
31.
ZhaoCShuXYanX, et al. RDD-YOLO: a modified YOLO for detection of steel surface defects. Measurement2023; 214: 112776.
32.
LuSWangJJingG, et al. Rail defect classification with deep learning method. Acta Polytech Hung2022; 19(6): 1–17.
33.
ZhaoJYeungAW-LAliM, et al. CBAM-SwinT-BL: small rail surface defect detection method based on Swin transformer with block level CBAM enhancement. IEEE Access2024; 12: 181997–182009.
34.
LuoHCaiLLiC.Rail surface defect detection based on an improved YOLOv5s. Appl Sci2023; 13(12): 7330.
35.
JiangYYanHZhangY, et al. RDD-YOLOv5: road defect detection algorithm with self-attention based on unmanned aerial vehicle inspection. Sensors2023; 23(19): 8241.
36.
YadavDPChauhanSKadaB, et al. Spatial attention-based dual stream transformer for concrete defect identification. Measurement2023; 218(3): 113137.
37.
LiuXPengHZhengN, et al. EfficientViT: memory efficient vision transformer with cascaded group attention. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, Vancouver, Canada, 17–24 June 2023, pp.14420–14430.
38.
HowardAGZhuMChenB, et al. Mobilenets: efficient convolutional neural networks for mobile vision applications. arXiv Preprint2017; 2017: arXiv:1704.04861.
39.
YangLZhangR-YLiL, et al. SimAM: A simple, parameter-free attention module for convolutional neural networks. Int Conf Mach Learn PMLR2021; 139: pp.11863–11874.
40.
TongZChenYXuZ, et al. Wise-IOU: bounding box regression loss with dynamic focusing mechanism. arXiv Preprint2023; 2023: arXiv:2301.10051.
41.
ÖzgenelÇFSorguçAG. Performance comparison of pretrained convolutional neural networks on crack detection in buildings. In: ISARC. Proceedings of the international symposium on automation and robotics in construction, IAARC Publications, Berlin, Germany, 20–25 July 2018, pp.1–8.
42.
YeWDengSRenJ, et al. Deep learning-based fast detection of apparent concrete crack in slab tracks with dilated convolution. Constr Build Mater2022; 329: 127157.
43.
HeYSongKMengQ, et al. An end-to-end steel surface defect detection approach via fusing multiple hierarchical features. IEEE Trans Instrum Meas2019; 69(4): 1493–1504.
44.
HeKZhangXRenS, et al. Deep residual learning for image recognition. In: Proceedings of the IEEE conference on computer vision and pattern recognition, Las Vegas, Nevada, 26 June–1 July 2016, pp.770–778.
45.
LiuZLinYCaoY, et al. Swin transformer: hierarchical vision transformer using shifted windows. In: Proceedings of the IEEE/CVF international conference on computer vision, Montreal, Canada, 11–17 October 2021, pp.10012–10022.
46.
SelvarajuRRCogswellMDasA, et al. Grad-cam: visual explanations from deep networks via gradient-based localization. In: Proceedings of the IEEE international conference on computer vision, Venice, Italy, 22–29 October 2017, pp.618–626.
