Research on Rail Damage Detection Based on Improved DETR Algorithm
Source abstract
Abstract
In rail damage detection, the scale variation of small targets leads to inaccurate extraction of damage morphology and size features, thereby affecting the reliable identification of damage types. The DETR algorithm has been optimized and improved. Firstly, we introduce the convolution–attention fusion module (CAFMAttention) after the two side convolutional layers of the original algorithm; then, we replace the nn.Upsample-based upsampling layer with the Dysample upsampler. Finally, we replace the Conv modules in the two down-sampled convolutional layers with Dual-Conv modules. The results of the comparative experiments show that the recall rate of the improved DETR model in this paper is 0.698, which is 12.2% higher than that of the original DETR model. The accuracy is 0.815, which is 2.3% higher than that of the original DETR model. The average precision (Map@0.5) is 0.741. Compared with the original DETR model, it has been improved by 8.7%. The F1 score is 0.75, which is 8.7% higher than the original DETR model. The frame per second (FPS) transfer rate is 64.94, which is 2.6% higher than that of the original DETR model. The proposed DETR algorithm can detect rail damage under complex working conditions well, with high accuracy and robustness, and better meet the requirements of practical actual rail detection.
Source: crossref · Quality: baseline
Research context
OpenAlex topics: Railway Engineering and Dynamics · Advanced Neural Network Applications · Advanced Data and IoT Technologies
SciVX discovery tags are terminology-based. OpenAlex topics are source-provided classifications.
Affiliations
School of Intelligent Manufacturing, Nanchang Jiaotong Institute, Nanchang 330013, China
Jiangxi Provincial Key Laboratory of Intelligent Operation and Maintenance Technology and Equipment for Rail Transit Vehicles, East China Jiaotong University, Nanchang 330013, China
Funding
Science and Technology Research Program of Jiangxi Provincial Department of Education · National Natural Science Foundation of China · Jiangxi Provincial Natural Science Foundation · Jiangxi Province Double Thousand Plan Science and Technology Innovation Leading Talent Project · Science and Technology Research and Development Program of China State Railway Group
References (46)
Show source-provided references
1. Rail Surface Damage Detection Method Based on Improved U-Net Convolutional Neural Network10.3788/LOP202158.0215009 ↗
2. Luo, H., Cai, L., and Li, C. (2023). Rail Surface Defect Detection Based on an Improved YOLOv5s. Appl. Sci., 13.10.3390/app13127330 ↗
3. Reliability Assessment of Railway Bridges Subjected to High-Speed Trains Considering the Effects of Seasonal Temperature Changes10.1016/j.engstruct.2016.08.017 ↗
4. Bai, T., Gao, J., Yang, J., and Yao, D. (2021). A Study on Railway Surface Defects Detection Based on an Improved YOLOv4. Entropy, 23.10.3390/e23111437 ↗
5. Du, J., Zhang, R., Gao, R., Nan, L., and Bao, Y. (2024). A New Multiscale Rail Surface Defect Detection Model (RSDNet) with YOLOv8n. Sensors, 24.10.3390/s24113579 ↗
6. Zhang, Y., Liu, X., Yang, Y., Li, N., Xie, X., and Chen, K. (2024). An Improved Target Network Model for Rail Surface Defect Detection Based on YOLOv7 and k-means++. Appl. Sci., 14.10.3390/app14156467 ↗
7. Rail Surface Defect Detection Using a Transformer-Based Network (RailFormer)
8. Si, C., Fan, R., Peng, F., Liang, H., and Chen, D. (2024). Rail-STrans: A Rail Surface Defect Segmentation Method Based on Swin Transformer. Appl. Sci., 14.10.3390/app14093629 ↗
9. Song, W., Liao, B., Ning, K., and Yan, X. (2024). Improved Real-Time Detection Transformer-Based Rail Fastener Defect Detection. Mathematics, 12.10.3390/math12213349 ↗
10. Real-Time Detection Method of Rail Corrugation Based on Machine Vision and CNN
11. Min, Y., Wang, Z., Liu, Y., and Wang, Z. (2023). FS-RSDD: Few-Shot Rail Surface Defect Detection with Prototype Learning. Sensors, 23.10.3390/s23187894 ↗
12. Rail Defect Detection Using Ultrasonic A-Scan Data and Deep Autoencoders10.1177/03611981221150923 ↗
13. Hybrid Physics–ML for Ultrasonic Field-Guided 3D Reconstruction of Rail Defects
14. Railway Track Surface Faults Dataset10.1016/j.dib.2024.110050 ↗
15. Zhang, Z., Yu, S., Yang, S., Zhou, Y., and Zhao, B. (2021). Rail-5k: A Real-World Dataset for Rail Surface Defects Detection. arXiv.
16. Qian, Y., Guo, H., Rizos, D., and Vitzilaios, N. (2024). Autonomous Rail Surface Defect Detection (RSD_UAV Dataset; DeepLabv3+ with CBAM). U.S. DOT/ROSA P Technical Report, U.S. DOT.
17. A Defect Detection Method for Rail Surface and Fasteners Based on Deep CNN10.1155/2021/2565500 ↗
18. Specificity Autocorrelation Integration Network (SAINet) for Surface Defect Detection of No-Service Rail10.1016/j.optlaseng.2023.107862 ↗
19. Mi, Z., Chen, R., and Zhao, S. (2023). Research on Steel Rail Surface Defects Detection Based on Improved YOLOv4 Network. Front. Neurorobotics, 17.10.3389/fnbot.2023.1119896 ↗
20. Attention Is All You Need
21. Self-Attention with Relative Position Representations
22. BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
23. Radford, A., Narasimhan, K., Salimans, T., and Sutskever, I. (2018). Improving Language Understanding by Generative Pre-Training. OpenAI Tech. Rep., 1–12.
24. XLNet: Generalized Autoregressive Pretraining for Language Understanding
25. Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer
26. Big Bird: Transformers for Longer Sequences
27. Ainslie, J., Ontañón, S., Alberti, C., Cvicek, V., Fisher, Z., Pham, P., Ravula, A., Sanghai, S., Wang, Q., and Yang, L. (2020). ETC: Encoding Long and Structured Inputs in Transformers. arXiv.10.18653/v1/2020.emnlp-main.19 ↗
28. Nawrot, P., Tworkowski, S., Tyrolski, M., Kaiser, Ł., Wu, Y., Szegedy, C., and Michalewski, H. (2021). Hierarchical Transformers Are More Efficient Language Models. arXiv.10.18653/v1/2022.findings-naacl.117 ↗
29. Schneider, J. (2024). What Comes After Transformers?—A Selective Survey. arXiv.
30. Pereira, F., and Hussain, S. (2024). Transformers in Computer Vision: A Comprehensive Survey. arXiv.
31. Zhu, X., Su, W., Lu, L., Li, B., Wang, X., and Dai, J. (2021, January 3–7). Deformable DETR: Deformable Transformers for End-to-End Object Detection. Proceedings of the International Conference on Learning Representations (ICLR), virtual.
32. Dai, X., Chen, Y., Xiao, B., Chen, D., Liu, M., Yuan, L., and Zhang, L. (2021, January 19–25). UP-DETR: Unsupervised Pre-Training for Object Detection with Transformers. Proceedings of the Conference on Computer Vision and Pattern Recognition, virtual.10.1109/CVPR46437.2021.00165 ↗
33. Gao, P., Zheng, M., Wang, X., Dai, J., and Li, H. (2021, January 11–17). Fast Convergence of DETR with Spatially Modulated Co-Attention. Proceedings of the International Conference on Computer Vision, virtual.10.1109/ICCV48922.2021.00360 ↗
34. Chen, H., Sun, K., Tian, Z., Shen, C., Huang, Y., and Yan, Y. (2022, January 19–24). Group DETR: Fast Training of DETR with Group-wise One-to-Many Assignment. Proceedings of the Conference on Computer Vision and Pattern Recognition, New Orleans, LA, USA.
35. Li, Y., Zhang, K., Zhou, H., Wang, Y., Jiang, Z., and Sun, J. (2023). Lite DETR: An Interleaved Multi-Scale Encoder for Efficient DETR. arXiv.10.1109/CVPR52729.2023.01780 ↗
36. Lv, T., Wang, W., Zhang, W., Yuan, Y., and Liu, J. (2023). RT-DETR: Real-Time Detection Transformer. arXiv.
37. Rank-DETR: Rank-Oriented DETR for High-Quality Object Detection
38. He, J., Zhang, Y., Chen, X., Wang, C., and Liu, J. (2024). An Improved DETR Detector for Oriented Remote Sensing Object Detection. Remote Sens., 16.10.3390/rs16183516 ↗
39. Nan, Z., Wang, J., Xu, Y., Li, B., and Wang, X. (2025, January 11–15). MI-DETR: An Object Detection Model with Multi-Time Inquiries Mechanism. Proceedings of the 2025 Conference on Computer Vision and Pattern Recognition, Nashville, TN, USA.10.1109/CVPR52734.2025.00443 ↗
40. Carion, N., Massa, F., Synnaeve, G., Usunier, N., Kirillov, A., and Zagoruyko, S. (2020, January 23–28). End-to-End Object Detection with Transformers. Proceedings of the Computer Vision—ECCV 2020, Glasgow, UK.10.1007/978-3-030-58452-8_13 ↗
41. Hybrid Convolutional and Attention Network for Hyperspectral Image Denoising10.1109/LGRS.2024.3489635 ↗
42. Imagrnent classification with deep convolutional neural networks
43. Singh, P., Verma, V.K., Rai, P., and Namboodiri, V.P. (2019, January 16–20). HetConv: Heterogeneous Kernel-Based Convolutions for Deep CNNs. Proceedings of the Conference on Computer Vision and Pattern Recognition, Long Beach, CA, USA.10.1109/CVPR.2019.00497 ↗
44. Dualconv: Dual convolutional kernels for lightweight deep neural networks10.1109/TNNLS.2022.3151138 ↗
45. Liu, W., Lu, H., and Fu, H. (2023, January 2–6). Learning to Upsample by Learning to Sample. Proceedings of the 2023 International Conference on Computer Vision, Paris, France.10.1109/ICCV51070.2023.00554 ↗
46. Pang, Y., Xie, J., Khan, M.H., Anwer, R.M., Khan, F.S., and Shao, L. (November, January 27). Mask-Guided Attention Network for Occluded Pedestrian Detection. Proceedings of the 2019 International Conference on Computer Vision, Seoul, Republic of Korea.10.1109/ICCV.2019.00507 ↗