Konvolüsyonel Blok Dikkat Modülü ile Entegre CNN Kullanılarak Verimli ve Gerçek Zamanlı Demiryolu Ray Arıza Sınıflandırması

Öz

Demiryolu ray yüzeylerinde çeşitli tipte arızaların oluşması tren raydan çıkması gibi kazalara yol açabilir. Bu çalışmada, ray arızalarının sınıflandırılması için evrişimsel sinir ağı ve evrişimsel blok dikkat modülünün birleştirilmesiyle çok hızlı ve sağlam sonuçlar üreten yeni bir yaklaşım geliştirilmiştir. Bu çalışmada kullanılan veri seti dört farklı arıza tipini içermekte olup, kamuya ait demiryolu veri seti üzerinde deneysel çalışmalar yürütülmüştür. Evrişimsel blok dikkat modülünün önerilen yaklaşımın performansına etkisi ve modelin genelleme yeteneğine katkısı incelenmiş ve önerilen yaklaşımın performansı, bu modül olmadan önerilen yaklaşımın performansına kıyasla yaklaşık %5 oranında artmaktadır. Önerilen yaklaşımın hızlı ve etkili sonuçlar üreterek demiryolu hat arıza teşhisinde etkin bir şekilde kullanılabileceği gösterilmiştir.

Anahtar Kelimeler

• Evrişimli blok dikkat modülü
• Dikkat mekanizması
• Ray arızası
• Ray yolu denetimi

Efficient and Real-Time Railway Track Fault Classification Using CNN Integrated with Convolutional Block Attention Module

Öz

The occurrence of various types of faults on railway rail surfaces can lead to accidents such as train derailments. In this study, a new approach that produces very fast and robust results is developed by combining the convolutional neural network and the convolutional block attention module for the classification of rail faults. The dataset used in this study contains four different fault types, and experimental studies were conducted on the public rail dataset. The impact of the convolutional block attention module on the performance of the proposed approach and its contribution to the model's generalization ability are examined, and the performance of the proposed approach increases by approximately 5% compared to the performance of the proposed approach without this module. It has been demonstrated that the proposed approach can be used effectively in railway track fault diagnosis by producing fast and effective results.

Anahtar Kelimeler

• Attention mechanism
• Convolutional block attention module
• Rail defect
• Rail track inspection

Kaynakça

1. [1] Y. Zhang, X. Wang, and Q. Liu, “Rail track fault detection using deep convolutional neural networks,” IEEE Access, vol. 8, pp. 130461–130470, 2020, doi: 10.1109/ACCESS.2020.3009424.
2. [2] L. Chen, H. Zhang, and P. Li, “Visual inspection in railway systems: A deep learning approach,” Sensors, vol. 19, no. 2, pp. 1–18, 2019, doi: 10.3390/s19020359.
3. [3] M. Faghih-Roohi, K. Cheng, R. Newman, et al., “Rail defect detection using ultrasonic technology and signal processing: A review,” Insight - Non-Destructive Testing and Condition Monitoring, vol. 56, no. 1, pp. 20–28, 2014.
4. [4] K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR), 2016, pp. 770–778, doi: 10.1109/CVPR.2016.90.
5. [5] C. Li, J. Zhao, Y. Liu, X. Li, and X. Han, "Rail surface defect detection based on deep learning," IEEE Transactions on Instrumentation and Measurement, vol. 70, pp. 1–14, 2021, doi: 10.1109/TIM.2021.3066373.
6. [6] W. Li, M. Chen, J. Liu, and Y. Xu, “Automated rail defect classification using attention-based deep networks,” Transp. Res. Part C Emerg. Technol., vol. 127, pp. 103115, 2021, doi: 10.1016/j.trc.2021.103115.
7. [7] S. A. Yadav, R. Tripathi, and M. Shukla, “Railway Track Fault Detection and Classification Using Hybrid Deep LearningTechniques,” IEEE Access, vol. 9, pp. 145892–145905, 2021.
8. [8] M. Z. Afzal, H. T. H. Nguyen, and Y. Kim, “Vision-based railway track monitoring using deep neural networks,” IEEE Sensors Journal, vol. 20, no. 23, pp. 14162–14170, 2020.

9. [9] F. J. Perez-Peña et al., “Real-time rail inspection system using computer vision,” Transportation Research Part C: Emerging Technologies, vol. 103, pp. 146–158, 2019.
10. [10] H. Tang, Z. Li, B. Wang, et al., “Deep learning-based surface defect detection for high-speed railway track inspection,” Neurocomputing, vol. 417, pp. 307–319, 2020.
11. [11] S. Woo, J. Park, J. Lee, and I. Kweon, “CBAM: Convolutional block attention module,” in Proc. Eur. Conf. Comput. Vis. (ECCV), 2018, pp. 3–19.
12. [12] Gao, B., Zhang, C., & Li, Y. (2020). "DeepRail: Railway defect classification using convolutional neural networks." Journal of Rail Transport Planning & Management, 16, 100221.
13. [13] Patel, V., Kumar, A., & Sharma, R. (2021). "Residual networks for track surface fault detection in Indian railways." Transportation Research Part C, 127, 102908.
14. [14] Lee, S., Choi, J., & Kim, D. (2019). "Rail defect classification using transfer learning on AlexNet." Sensors, 19(21), 4751.
15. [15] Park, D., Yoon, S., & Lee, B. (2022). "Real-time lightweight CNN for railway surface defect classification." IEEE Sensors Journal, 22(5), 4102–4110.
16. [16] Rao, H., & Singh, M. (2018). "Multi-class defect detection on railway track using VGGNet and augmentation." Procedia Computer Science, 132, 785–792.
17. [17] Zhang, X., Liu, Y., Wang, H. (2021). “Deep learning-based railway surface defect classification with attention mechanisms,” IEEE Access, vol. 9, pp. 125441–125452.
18. [18] Li, J., Chen, W., & Wu, Y. (2022). “Multiscale attention network for rail surface defect detection,” Sensors, 22(4), 1547.
19. [19] Wang, M., Lu, Q., & Liu, H. (2021). “EfficientNet and CBAM for railway track damage classification,” Transportation Research Part C, 132, 103391.
20. [20] Park, D., Choi, H., Kim, Y. (2023). “Rail track anomaly detection via GAF-based image analysis and CBAM-enhanced CNNs,” Engineering Applications of Artificial Intelligence, 123, 105558.
21. [21] Song, K., & Yan, Y. (2021). “A multi-attention deep learning model for rail surface defect classification,” Measurement, 178, 109376.
22. [22] Zhou, X., Wang, Q., & Zhang, Y. (2022). “Lightweight attention networks for surface defect classification in rail inspection,” Neural Computing and Applications, 34(7), 5617–5632.
23. [23] Bakır, F., Aksoy, F., & Demir, C. (2023). “Deep Attention Mechanisms for Track Surface Defect Detection,” Expert Systems with Applications, 208, 118202.
24. [24] Aydin, I., Akin, E., & Karakose, M. (2021). Defect classification based on deep features for railway tracks in sustainable transportation. Applied Soft Computing, 107706.