基于改进YOLOv5的超声钢轨缺陷检测方法
Ultrasonic Rail Defect Target Detection Based on Improved YOLOv5
DOI: 10.12677/ORF.2023.132138, PDF,    科研立项经费支持
作者: 任恩璇, 汪 磊：上海工程技术大学，城市轨道交通学院，上海；鲍蟠虎, 王 振：江苏三合声源超声波科技有限公司，江苏 常州
关键词: 钢轨缺陷检测；注意力机制；YOLOv5；超声相控阵图像；Rail Defect Detection； Attention Mechanism； YOLOv5； Ultrasonic Phased Array Image
摘要: 钢轨内部缺陷检测是保证列车运行安全非常重要的安全保证，为了增加对钢轨内部缺陷的识别准确率，减少人为不确定因素的干扰，提出了一种用YOLOv5目标识别的方法来辅助识别，并对YOLOv5算法进行改进，改进了数据增强，提升数据鲁棒性，添加NAM注意力模块来捕获局部信息，同时引入GhostNet，对模型进行轻量化处理。改进的YOLOv5_GNM算法的AP50值到达了87.7%，参数量减少了37%，帧数达到了66帧，提高了缺陷识别的准确率。
Abstract: Rail internal defect detection is a very important safety guarantee to ensure the safety of train operation. In order to increase the recognition accuracy of rail internal defects and reduce the interference of human uncertainties, a YOLOv5 target recognition method is proposed to assist the recognition, and YOLOv5 algorithm is improved to improve data enhancement, improve data robustness, add NAM attention module to capture local information, introduce GhostNet, and lightweight the model. The AP50 value of the improved YOLOv5_GNM algorithm reaches 87.7%, the number of parameters decreases by 37%, and the number of frames reaches 66, which im-proves the accuracy of defect recognition.
文章引用：任恩璇, 汪磊, 鲍蟠虎, 王振. 基于改进YOLOv5的超声钢轨缺陷检测方法[J]. 运筹与模糊学, 2023, 13(2): 1362-1371. https://doi.org/10.12677/ORF.2023.132138

参考文献

[1] 张学强. 超声相控阵探伤中缺陷图谱处理与智能分类技术的研究[D]: [硕士学位论文]. 包头: 内蒙古科技大学, 2021.
[2] 薛文杰. 超声相控阵全聚焦成像实现与降噪方法研究[D]: [硕士学位论文]. 南京: 东南大学, 2021.
[3] Kong, X.X. and Li, J. (2018) Image Registration-Based Bolt Loosening Detection of Steel Joints. Sensors (Basel, Switzerland), 18, 1000. [Google Scholar] [CrossRef] [PubMed]
[4] Girshick, R., Donahue, J., Darrell, T., et al. (2014) Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation. Pro-ceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Columbus, 23-28 June 2014, 580-587. [Google Scholar] [CrossRef] 
[5] Girshick, R. (2015) Fast R-CNN. Proceedings of the IEEE International Conference on Computer Vision, Santiago, 7-13 December 2015, 1440-1448. [Google Scholar] [CrossRef] 
[6] Ren, S., He, K., Girshick, R., et al. (2017) Faster R-CNN: To-wards Real-Time Object Detection with Region Proposal Networks. IEEE Transactions on Pattern Analysis & Machine Intelligence, 39, 1137-1149. [Google Scholar] [CrossRef] 
[7] Liu, W., Anguelov, D., Erhan, D., et al. (2016) Ssd: Sin-gleshot Multibox Detector. In: European Conference on Computer Vision, Springer, Cham, 21-37. [Google Scholar] [CrossRef] 
[8] Redmon, J., Divvala, S., Girshick, R., et al. (2016) You Only Look Once: Unified, Real-Time Object Detection. IEEE Conference on Computer Vision and Pattern Recog-nition, Las Vegas, 27-30 June 2016, 779-788. [Google Scholar] [CrossRef] 
[9] Redmon, J. and Farhadia, A. (2017) YOLO9000: Better, Faster, Stronger. Proceedings of 2017 IEEE Conference on Computer Vision and Pattern Recognition, Honolulu, 21-26 July 2017, 6517-6525. [Google Scholar] [CrossRef] 
[10] Joseph, R. and Ali, F. (2018) YOLOv3: An Incremental Im-provement.
[11] Bochkovskiy, A., Wang, C.Y. and Liao, H.Y.M. (2020) YOLOv4: Optimal Speed and Accuracy of Object Detection.
[12] 仇娇慧, 贝绍轶, 尹明锋, 卿宏军. 基于改进YOLOv5s的齿轮表面缺陷检测[J]. 现代制造工程, 2022(3): 104-113. [Google Scholar] [CrossRef] 
[13] 王淑青, 张鹏飞, 要若天, 鲁东林, 顿伟超. 基于改进YOLOv5的太阳能电池片表面缺陷检测[J]. 仪表技术与传感器, 2022(5): 111-116.
[14] 费学智, 吴军芳, 柳铭哲, 等. 超声相控阵图像中的缺陷识别算法[J]. 无损检测, 2018, 40(8): 61-66.
[15] 吉春生. 基于改进Faster R-CNN的焊缝缺陷图谱识别[J]. 中国化工装备, 2022, 24(2): 26-32+36.
[16] Liu, Y., Shao, Z., Teng, Y. and Hoffmann, N. (2021) NAM: Normalization-Based Attention Module.
[17] Han, K., Wang, Y.H., Tian, Q., et al. (2020) GhostNet: More Features from Cheap Operations. Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Seattle, 13-19 June 2020, 1577-1586. [Google Scholar] [CrossRef] 

为你推荐