Aiming at the problem of insufficient feature information of remote and small target images in railway long-depth monitoring scenarios, investigation on the video super-resolution reconstruction technology is conducted. Firstly, based on the BasicVSR network, the Feature Interaction Enhancement (FIE) module and the Graph Laplacian Pyramid Low-Frequency Separation (GLPLS) module are designed relying on the information from preceding and succeeding frames, thereby constructing an efficient feature reconstruction network RailVSR of remote and small targets. Secondly, a joint loss function is integrated into the RailVSR network to optimize the dual attention of network on high-resolution image quality and target detection precision. Lastly, by combining RailVSR network with the RT-DETR target detection algorithm, the capability for detecting remote and small target intrusions in the long-depth monitoring scenarios of railway is enhanced. The results demonstrate that compared with the original RT-DETR target detection algorithm, the improved algorithm based on the RailVSR network at least achieves a 13% improvement in the detection precision of railway perimeter intrusions, with an average precision increase of at least 11%. In the railway monitoring dataset constructed by the VSTR railway sample database, both the detection false negative rate and false positive rate are 0 when the proportion of target pixel exceeds 0.05%, and the average detection precision can reach over 85%. These findings confirm the effectiveness of the proposed method in intrusion detection of remote and small targets along railway perimeters, and significantly enhance the safety of railway operations.
DONGHonghui, GEDawei, QINYong, et al. Research on Railway Invasion Detection Technology Based on Intelligent Video Analysis [J]. China Railway Science, 2010, 31 (2): 121-125. in Chinese
GUANLing, JIALimin, XIEZhengyu. Research on Lightweight Model for Railway Intrusion Detection Integrating Attention Mechanism [J]. Journal of the China Railway Society, 2023, 45 (5): 72-81. in Chinese
[5]
HARISM, SHAKHNAROVICHG, UKITAN. Task-Driven Super Resolution: Object Detection in Low-Resolution Images [C]// Neural Information Processing. Cham: Springer International Publishing, 2021: 387-395.
[6]
JINY, ZHANGY, CENY G, et al. Pedestrian Detection with Super-Resolution Reconstruction for Low-Quality Image [J]. Pattern Recognition, 2021, 115: 107846.
ZHULiqiang, XULizhi, ZHAOWenyu, et al. Multi-Scale Feature Perception Algorithm for Railway Perimeter Intrusion Object [J]. China Railway Science, 2024, 45 (1): 215-226. in Chinese
WANGRuifeng, CHENXiaoyi. Research on Orbital Foreign Object Intrusion Detection Algorithm Based on Improved YOLOv5 [J]. Journal of Yunnan University (Natural Sciences Edition), 2023, 45 (4): 799-806. in Chinese
HUHao, SHITianyun, YANGWen. Railway Perimeter Intrusion Detection Method Based on Improved FairMOT [J]. China Railway Science, 2023, 44 (5): 222-232. in Chinese
CAIXinnan, DINGXuewen, ZHANGZiyi, et al. Intrusion Detection of Rail Foreign Objects Based on YOLOv5 [J]. Intelligent Computer and Applications, 2022, 12 (11): 161-167. in Chinese
ZHANGJian, WANGDengzhun, MOGuangjian, et al. High-Speed Rail Foreign Body Intrusion Detection Algorithm Based on Improved YOLOv3 [J]. Computer Technology and Development, 2022, 32 (2): 69-74. in Chinese
XUXin, PANJie, CAOLi’an, et al. Railway Foreign Object Intrusion Detection Model Based on Deep Learning [J]. Railway Computer Application, 2023, 32 (10): 7-12. in Chinese
[19]
LIUX B, CHENS Q, SONGL P, et al. Self-Attention Negative Feedback Network for Real-Time Image Super-Resolution [J]. Journal of King Saud University - Computer and Information Sciences, 2022, 34 (8): 6179-6186.
[20]
ZAMIRS W, ARORAA, KHANS, et al. Restormer: Efficient Transformer for High-Resolution Image Restoration [C]//2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). New Orleans, LA, USA. New York: IEEE Press, 2022: 5718-5729.
[21]
ZHAOX T, LIW, ZHANGY F, et al. Residual Super-Resolution Single Shot Network for Low-Resolution Object Detection [J]. IEEE Access, 2018, 6: 47780-47793.
[22]
GUOJ, LAIS C, TUQ H, et al. Ultra-High Resolution SVBRDF Recovery from a Single Image [J]. ACM Transactions on Graphics, 2023, 42 (3): 1-14.
[23]
ZHOUL G, CHENG, FENGM Y, et al. Improving Low-Resolution Image Classification by Super-Resolution with Enhancing High-Frequency Content [C]// 2020 25th International Conference on Pattern Recognition (ICPR). Milan, Italy. New York: IEEE Press, 2021: 1972-1978.
[24]
刘义健. 基于图像超分辨率重建算法的轨道异物入侵检测方法[D]. 兰州:兰州交通大学,2021.
[25]
LIUYijian. Research on Orbital Foreign Object Intrusion Detection Based on Image Super-Resolution Reconstruction [D]. Lanzhou: Lanzhou Jiaotong University, 2021. in Chinese
ZHAOBing. Research and Application of Key Technologies of Railway Image intelligent Analysis Based on Deep Learning [D]. Beijing: China Academy of Railway Sciences, 2020. in Chinese
[28]
ZHANGD Y, PUH J, LIF, et al. Few-Shot Object Detection Based on the Transformer and High-Resolution Network [J]. Computers, Materials & Continua, 2023, 74 (2): 3439-3454.
[29]
ZHANGX W, LIUQ Y, CHANGH L, et al. High-Resolution Network with Transformer Embedding Parallel Detection for Small Object Detection in Optical Remote Sensing Images [J]. Remote Sensing, 2023, 15 (18): 4497.
[30]
RONNEBERGERO, FISCHERP, BROXT. U-Net: Convolutional Networks for Biomedical Image Segmentation [C]// Medical Image Computing and Computer-Assisted Intervention-MICCAI 2015. Cham: Springer International Publishing, 2015: 234-241.
Zhongguancun Rail Transit Video and Security Industry Technology Alliance for Rail Transit. VSTR Railway Perimeter Intrusion Target Sample Database [EB/OL]. (2018-11-11) [2024-06-16] in Chinese
[35]
GUOK H, ZHANGZ, GUOH F, et al. Video Super-Resolution Based on Inter-Frame Information Utilization for Intelligent Transportation [J]. IEEE Transactions on Intelligent Transportation Systems, 2023, 24 (11): 13409-13421.
[36]
CHANK C K, WANGX T, YUK, et al. BasicVSR: the Search for Essential Components in Video Super-Resolution and beyond [C]// 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Nashville, TN, USA. New York: IEEE Press, 2021: 4945-4954.
[37]
CHANK C K, ZHOUS C, XUX Y, et al. Investigating Tradeoffs in Real-World Video Super-Resolution [C]// 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). New Orleans, LA, USA. New York: IEEE Press, 2022: 5952-5961.
[38]
CHANK C K, ZHOUS C, XUX Y, et al. BasicVSR++: Improving Video Super-Resolution with Enhanced Propagation and Alignment [C]// 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). New Orleans, LA, USA. New York: IEEE Press, 2022: 5962-5971.