复杂情况下列车轮对踏面缺陷检测研究

doi:10.6040/j.issn.1672-3961.0.2025.090

摘要/Abstract

摘要： 针对复杂情况下列车轮对踏面缺陷检测效果差、计算量、参数量与精度不匹配的问题,提出一种基于改进YOLOv9s的检测方法。用稀疏性网络结构Ghost-Inception替代YOLOv9s骨干网络中的RepNCSPELAN4,降低模型参数量与计算量,提升检测精度和效率;在Neck网络引入iEMA和SEAM注意力机制,新建PagAM特征拼接模块,增强复杂情况下的特征提取能力;构建RepNCSPELAN4-DCNv3特征提取-融合模块,加强对不同尺度缺陷的检测。试验结果表明,改进的YOLOv9s网络参数量减少8.9%,计算量降低11.4%,平均精度均值提升5.0%。该方法在算法轻量化前提下实现检测精度提升。

关键词: 列车轮对踏面, 缺陷检测, YOLOv9s, 轻量化, 精度提升

Abstract: To address the problems of poor detection performance for train wheelset tread defects under complex conditions and the mismatch among computational load, parameter count and accuracy, a detection method based on improved YOLOv9s was proposed. The sparse network structure Ghost-Inception was used to replace RepNCSPELAN4 in the backbone network of YOLOv9s, so as to reduce the model parameters and computational load, and improve detection accuracy and efficiency. The iEMA and SEAM attention mechanisms were introduced into the Neck network, and a new PagAM feature concatenation module was constructed to enhance the feature extraction capability under complex conditions. A RepNCSPELAN4-DCNv3 feature extraction and fusion module was built to strengthen the detection of defects at different scales. Experimental results showed that the parameters of the improved YOLOv9s network were reduced by 8.9%, the computational load was decreased by 11.4%, and the mean average precision was increased by 5.0%. The proposed method achieved improved detection accuracy on the premise of algorithm lightweight.

Key words: train wheelset tread, defect detection, YOLOv9s, lightweight, precision improvement

中图分类号:

TP391

侯涛,任祎坤,牛宏侠. 复杂情况下列车轮对踏面缺陷检测研究[J]. 山东大学学报 (工学版), 2026, 56(5): 1-12.

Hou Tao, Ren Yikun, Niu Hongxia. Research on the detection of train wheelset tread defects under complex conditions[J]. Journal of Shandong University(Engineering Science), 2026, 56(5): 1-12.

参考文献

[1] 肖乾, 姜雄峰, 刘海涛, 等. 铁道车轮踏面损伤实时监测方法研究综述[J]. 华东交通大学学报, 2021, 38(4): 99-112. Xiao Qian, Jiang Xiongfeng, Liu Haitao, et al. Research on real-time monitoring methods for railroad wheel tread defects: a review[J]. Journal of East China Jiaotong University, 2021, 38(4): 99-112.
[2] 刘应桃. 轮对踏面检测系统设计与缺陷智能识别技术研究[D]. 成都: 西南交通大学, 2023: 1. Liu Yingtao. Research on wheel-to-tread inspection system design and defect intelligent identification technology[D]. Chengdu: Southwest Jiaotong University, 2023: 1.
[3] Ye Y G, Li H Q, Wang Q S, et al. Fault diagnosis of railway wheelsets: a review[J]. Measurement, 2025, 242: 116169.
[4] He K L, Peng J P, Ma R Y, et al. Dynamic detection of wheel tread crack based on eddy current pulse thermography[C] //2021 IEEE Far East NDT New Technology & Application Forum. Kunming, China: IEEE, 2021: 93-98.
[5] Zhang C F, Huang C C, He J. Defects recognition of train wheelset tread based on improved spiking neural network[J]. Chinese Journal of Electronics, 2023, 32(5): 941-954.
[6] Shangguan A Q, Xie G, Fei R, et al. Train wheel degradation generation and prediction based on the time series generation adversarial network[J]. Reliability Engineering & System Safety, 2023, 229: 108816.
[7] 刘二林, 刘成刚, 姜香菊, 等. 基于灰度-梯度共生矩阵的车轮踏面缺陷聚类分析[J]. 光电子·激光, 2022, 33(1): 53-60. Liu Erlin, Liu Chenggang, Jiang Xiangju, et al. Cluster analysis of wheel tread defects based on gray-gradient cooccurrence matrix[J]. Journal of Optoelectronics·Laser, 2022, 33(1): 53-60.
[8] 张昌凡, 徐逸夫, 何静, 等. 基于残差注意力的YOLO-v5列车轮对踏面缺陷快速检测[J]. 机车电传动, 2022(6): 1-9. Zhang Changfan, Xu Yifu, He Jing, et al. Rapid detection of wheel tread defects for YOLO-v5 trains based on residual attention[J]. Electric Drive for Locomotives, 2022(6): 1-9.
[9] 刘应桃, 郭世伟, 付孟新, 等. 基于改进Faster RCNN的轮对踏面缺陷检测[J]. 电子测量技术, 2023, 46(12): 34 -41. Liu Yingtao, Guo Shiwei, Fu Mengxin, et al. Defect detection of wheelset tread based on improved Faster RCNN[J]. Electronic Measurement Technology, 2023, 46(12): 34-41.
[10] 钱尚乐, 曹伟, 高军伟. 基于改进YOLOv7-tiny的轻量化列车轮对踏面缺陷检测方法[J]. 光电子·激光, 2025, 36(7): 733-744. Qian Shangle, Cao Wei, Gao Junwei. Lightweight train wheelset tread defect detection method based on improved YOLOv7-tiny[J]. Journal of Optoelectronics·Laser, 2025, 36(7): 733-744.
[11] Wang C Y, Yen I H, Mark Liao H Y. YOLOv9: learning what you want to learn using programmable gradient information[C] //Computer Vision-ECCV 2024. Milano, Italy: Springer, 2024: 1-21.
[12] Szegedy C, Vanhoucke V, Ioffe S, et al. Rethinking the inception architecture for computer vision[C] // 2016 IEEE Conference on Computer Vision and Pattern Recognition(CVPR). Las Vegas, USA: IEEE, 2016: 2818-2826.
[13] Szegedy C, Liu W, Jia Y, et al. Going deeper with convolutions[C] // 2015 IEEE Conference on Computer Vision and Pattern Recognition(CVPR). Boston, USA: IEEE, 2015: 1-9.
[14] Han K, Wang Y H, Tian Q, et al. GhostNet: more features from cheap operations[C] // 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR). Seattle, USA: IEEE, 2020: 1577-1586.
[15] Xu J C, Xiong Z X, Bhattacharyya S P. PIDNet: a real-time semantic segmentation network inspired by PID controllers[C] // 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR). Vancouver, Canada: IEEE, 2023: 19529-19539.
[16] Yang L X, Zhang R Y, Li L D, et al. SimAM: a simple, parameter-free attention module for convolutional neural networks[C] // International Conference on Machine Learning. [S.l.] : PMLR, 2021: 11863-11874.
[17] Ouyang D L, He S, Zhang G Z, et al. Efficient multi-scale attention module with cross-spatial learning[C] // ICASSP 2023-2023 IEEE International Conference on Acoustics, Speech and Signal Processing(ICASSP). Rhodes Island, Greece: IEEE, 2023: 1-5.
[18] Zhang J N, Li X T, Li J, et al. Rethinking mobile block for efficient attention-based models[C] // 2023 IEEE/CVF International Conference on Computer Vision(ICCV). Paris, France: IEEE, 2023: 1389-1400.
[19] Yu Z P, Huang H B, Chen W J, et al. YOLO-FaceV2: a scale and occlusion aware face detector[J]. Pattern Recognition, 2024, 155: 110714.
[20] Wang W H, Dai J F, Chen Z, et al. InternImage: exploring large-scale vision foundation models with deformable convolutions[C] // 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR). Vancouver, Canada: IEEE, 2023: 14408-14419.
[21] Zhu X Z, Hu H, Lin S, et al. Deformable ConvNets V2: more deformable, better results[C] // 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR). Long Beach, USA: IEEE, 2019: 9308-9316.
[22] Wang A, Chen H, Lin Z J, et al. Rep ViT: revisiting mobile CNN from ViT perspective[C] // 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR). Seattle, USA: IEEE, 2024: 15909-15920.
[23] Song Y D, He Z Q, Qian H, et al. Vision transformers for single image dehazing[J]. IEEE Transactions on Image Processing, 2023, 32: 1927-1941.
[24] Chen Z X, He Z W, Lu Z M. DEA-net: single image dehazing based on detail-enhanced convolution and content-guided attention[J]. IEEE Transactions on Image Processing, 2024, 33: 1002-1015.

多维度评价

Viewed

Full text

Abstract

Cited

Shared

Discussed