Hybridrailnet: A 5g-Enabled Dual-Stream Fmcw Radar Framework For Real Time Railway Obstacle Detection
DOI:
https://doi.org/10.29284/7j1j7q93Keywords:
Railway obstacle detection, FMCW radar, Range–Doppler, Micro-Doppler, Dual-stream CNN, Attention mechanism.Abstract
Real-time obstacle identification under low light, fog, rain, and strong rail-clutter reflections is necessary to prevent safety-critical dangers caused by pedestrian, animal, and vehicle intrusions on railway tracks. While single-stream lightweight CNNs frequently experience noisy Range– Doppler patterns, limited motion discrimination, and higher false alarms/missed detections, conventional vision/thermal systems deteriorate in low visibility, and classical radar pipelines (hand-crafted CFAR/thresholding + rule-based logic) are brittle under nonstationary clutter and multipath. This study provide a 5G-enabled IoT radar obstacle recognition system that uses HybridRailNet, a dual-stream CNN that simultaneously learns from Range–Doppler (RD) and Micro-Doppler (MD) representations, to overcome these drawbacks. After applying Min–Max normalization, MTI-based clutter suppression, and a CFAR-like adaptive mask, the pipeline uses an enhanced ResNet-18 backbone with asymmetric convolution fusion, Mish activation, and Improved CBAM (ICBAM) attention to process RD and MD, followed by late fusion and Focal Loss training. HybridRailNet with Accuracy (0.9889), Precision(1.0000), Recall(0.9789) and F1- Score(0.9893). With a low bandwidth need (0.008), it achieves 4.2 ms edge latency and 14.7 ms end-to-end alert latency for 5G IoT applicability. These findings show that HybridRailNet supports reliable 5G-integrated safety monitoring by enabling precise, low-latency, and communication- efficient railway obstacle notifications.
References
[1] Y. Liang, X. Qian, and L. Zhu, “Towards better railway service: Passengers counting in railway compartment,” IEEE Trans. Circuits Syst. Video Technol., vol. 31, no. 2, pp. 439– 451, 2020.
[2] T. Ye, B. Wang, P. Song, and J. Li, “Automatic railway traffic object detection system using feature fusion refine neural network under shunting mode,” Sensors, vol. 18, no. 6, p. 1916, 2018.
[3] J. Vazquez et al., “Detection acof moving objects in railway using vision,” in IEEE Intelligent Vehicles Symposium, 2004, IEEE, 2004, pp. 872–875.
[4] R. Kapoor, R. Goel, and A. Sharma, “An intelligent railway surveillance framework based on recognition of object and railway track using deep learning,” Multimed. Tools Appl., vol. 81, no. 15, pp. 21083–21109, 2022.
[5] Z. Zhao, J. Kang, Z. Sun, T. Ye, and B. Wu, “A real-time and high-accuracy railway obstacle detection method using lightweight CNN and improved transformer,” Measurement, vol. 238, p. 115380, 2024.
[6] P. V Vimal and K. S. Shivaprakasha, “IOT based greenhouse environment monitoring and controlling system using Arduino platform,” in 2017 International Conference on Intelligent Computing, Instrumentation and Control Technologies (ICICICT), IEEE, 2017,pp. 1514–1519.
[7] M. S. Adhikari, R. Patel, M. Sindhwani, and S. Sachdeva, “Introduction to AI Techniques for 6G Application,” Dev. 6G Networks Technol., pp. 1–27, 2024.
[8] M. Bakyt, L. La Spada, N. Zeeshan, K. Moldamurat, and S. Atanov, “Application of Quantum Key Distribution to Enhance Data Security in Agrotechnical Monitoring Systems Using UAVs.,” Appl. Sci., vol. 15, no. 5, 2025.
[9] Y. Ahmed, M. A. Azad, and T. Asyhari, “Rapid forecasting of cyber events using machine learning-enabled featureacs,” Information, vol. 15, no. 1, p. 36, 2024.
[10] A. Bhattacharyya, S. M. Nambiar, R. Ojha, A. Gyaneshwar, U. Chadha, and K. Srinivasan, “Machine Learning and Deep Learning powered satellite communications: Enabling technologies, applications, open challenges, and future research directions,” Int. J. Satell. Commun. Netw., vol. 41, no. 6, pp. 539–588, 2023.
[11] B. Akoramurthy and B. Surendiran, “Analysis of Techniques and Methods for Health Informatics: A Quantum Leap,” in Intelligent Computing Systems and Applications: Proceedings of the 2nd International Conference, ICICSA 2023, Springer Nature, 2024, p. 261.
[12] A. Jabbar, H. Jianjun, M. K. Jabbar, K. ur Rehman, and T. Mahmood, “Quantum-assisted federated learning for radar-based object tracking in IoT-enabled environments,” EPJ Quantum Technol., vol. 12, no. 1, p. 141, 2025.
[13] C. De Alwis, Q.-V. Pham, and M. Liyanage, 6G Frontiers: Towards Future Wireless Systems. John Wiley & Sons, 2022.
[14] A. Krizhevsky, I. Sutskever, and G. E. Hinton, “ImageNet Classification with Deep Convolutional Neural Networks,” in Advances in Neural Information Processing Systems, F. Pereira, C. J. Burges, L. Bottou, and K. Q. Weinberger, Eds., Curran Associates, Inc., 2012.[Online].Available:
https://proceedings.neurips.cc/paper_files/paper/2012/file/c399862d3b9d6b76c8436e924a 68c45b-Paper.pdf
[15] J. Redmon, S. Divvala, R. Girshick, and A. Farhadi, “You only look once: Unified, real- time object detection,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 779–788.
[16] H. Pan, Y. Li, H. Wang, and X. Tian, “Railway obstacle intrusion detection based on convolution neural network multitask learning,” Electronics, vol. 11, no. 17, p. 2697, 2022.
[17] W. Jiang, Y. Wang, Y. Li, Y. Lin, and W. Shen, “Radar target characterization and deep learning in radar automatic target recognition: A review,” Remote Sens., vol. 15, no. 15, p. 3742, 2023.
[18] J. Redmon and A. Farhadi, “YOLOv3: An Incremental Improvement,” 2018, [Online]. Available: http://arxiv.org/abs/1804.02767
[19] A. Bochkovskiy, C.-Y. Wang, and H.-Y. M. Liao, “YOLOv4: Optimal Speed and Accuracy of Object Detection,” 2020, [Online]. Available: http://arxiv.org/abs/2004.10934
[20] W. Liu 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.
[21] T.-Y. Lin, P. Goyal, R. Girshick, K. He, and P. Dollár, “Focal loss for dense object detection,” in Proceedings of the IEEE international conference on computer vision, 2017,pp. 2980–2988.
[22] R. Girshick, J. Donahue, T. Darrell, and J. Malik, “Region-based convolutional networks for accurate object detection and segmentation,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 38, no. 1, pp. 142–158, 2015.
[23] S. Ren, K. He, R. Girshick, and J. Sun, “Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks,” in Advances in Neural Information Processing Systems, C. Cortes, N. Lawrence, D. Lee, M. Sugiyama, and R. Garnett, Eds., Curran Associates, Inc., 2015. [Online]. Available:
https://proceedings.neurips.cc/paper_files/paper/2015/file/14bfa6bb14875e45bba028a21e d38046-Paper.pdf
[24] H. Tabassum, J. Jenita, S. Hunagund, S. Sahana, S. S. Santolli, and S. A. Rehman, “IoT- enabled Asset Monitoring and Optimization Railway System,” in 2024 IEEE 3rd World Conference on Applied Intelligence and Computing (AIC), IEEE, 2024, pp. 1242–1248.
[25] X. Zhou, C. Yang, and W. Yu, “Moving object detection by detecting contiguous outliers in the low-rank representation,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 35, no. 3, pp. 597–610, 2012.
[26] R. Manikandan, M. Balasubramanian, and S. Palanivel, “Vision based obstacle detection on railway track,” Int. J. Pure Appl. Math, vol. 116, no. 24, pp. 567–576, 2017.
[27] G. Krummenacher, C. S. Ong, S. Koller, S. Kobayashi, and J. M. Buhmann, “Wheel defect detection with machine learning,” IEEE Trans. Intell. Transp. Syst., vol. 19, no. 4, pp. 1176– 1187, 2017.
[28] L. Tong, L. Zhu, Z. Yu, and B. Guo, “Railway obstacle detection using onboard forward- viewing camera,” J Transp Syst Eng Inf Technol, vol. 12, no. 4, pp. 79–83, 2012.
[29] S. Mangale, R. Tambe, and M. Khambete, “OBJECT DETECTION AND TRACKING IN THERMAL VIDEO USING DIRECTED ACYCLIC GRAPH (DAG).,” ICTACT J. Image Video Process., vol. 8, no. 1, 2017.
[30] M. Krišto, M. Ivasic-Kos, and M. Pobar, “Thermal object detection in difficult weather conditions using YOLO,” IEEE access, vol. 8, pp. 125459–125476, 2020.
[31] D. He, Z. Zou, Y. Chen, B. Liu, and J. Miao, “Rail transit obstacle detection based on improved CNN,” IEEE Trans. Instrum. Meas., vol. 70, pp. 1–14, 2021.
[32] Q. Guo, L. Liu, W. Xu, Y. Gong, X. Zhang, and W. Jing, “An improved faster R-CNN for high-speed railway dropper detection,” IEEE Access, vol. 8, pp. 105622–105633, 2020.
[33] D. He et al., “Improved Mask R-CNN for obstacle detection of rail transit,” Measurement, vol. 190, p. 110728, 2022.
[34] Y. Xu, H. Tao, and L. Hu, “Research on foreign object encroachment detection algorithm for high-speed railroad based on faster R-CNN,” J China Railw Soc, vol. 42, no. 05, pp. 91– 98, 2020.
[35] H. Niu, D. Feng, and T. Hou, “Research on foreign object intrusion detection in railway tracks based on MSL-YOLO,” J. Eng. Appl. Sci., vol. 72, no. 1, p. 136, 2025.
[36] F. U. Rahman, M. T. Ahmed, M. M. Hasan, and N. Jahan, “Real-Time Obstacle Detection Over Railway Track using Deep Neural Networks,” Procedia Comput. Sci., vol. 215, pp. 289–298, 2022, doi: https://doi.org/10.1016/j.procs.2022.12.031.
[37] S. Liu and J. Lin, “MRS-YOLO Railroad Transmission Line Foreign Object Detection Based on Improved YOLO11 and Channel Pruning,” arXiv Prepr. arXiv2510.10553, 2025.
[38] B. Song, S. Zhao, Z. Wang, J. Chen, W. Liu, and X. Liu, “LW-DETR: a lightweight transformer-based object detection algorithm for efficient railway crossing surveillance,” Complex Intell. Syst., vol. 11, no. 12, pp. 1–13, 2025.
[39] Z. Qi, D. Ma, J. Xu, A. Xiang, and H. Qu, “Improved YOLOv5 based on the attention mechanism and FasterNet for foreign object detection on railway and airway tracks,” in 2024 Asian Conference on Communication and Networks (ASIANComNet), IEEE, 2024, pp. 1–6.
[40] Z. Dong, Q. Yang, H. Chen, H. Zhou, and D. Gao, “A lightweight transformer-based framework for real-time foreign object detection in complex railway environments,” J. Real-Time Image Process., vol. 23, Nov. 2025, doi: 10.1007/s11554-025-01800-0.
[41] V. Suresh, P. Janik, J. Rezmer, and Z. Leonowicz, “Forecasting solar PV output using convolutional neural networks with a sliding window algorithm,” Energies, vol. 13, no. 3, p. 723, 2020.
[42] K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 770–778.
[43] J. Park, S. Woo, J.-Y. Lee, and I. S. Kweon, “Bam: Bottleneck attention module,” arXiv Prepr. arXiv1807.06514, 2018.
Downloads
Published
Issue
Section
License
This work is licensed under a Creative Commons Attribution 4.0 International License.
This work is licensed under a Creative Commons Attribution 4.0 International License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
