|
|
|
| Rail Surface Defect Detection Based on Improved YOLOv7 |
| CHEN Renxiang1, PAN Sheng1, YANG Lixia2, GAO Xiaopeng3, WANG Jianxi4 |
1. Chongqing Engineering Laboratory for Transportation Engineering Application Robot, Chongqing Jiaotong University, Chongqing 400074, China;
2. Chongqing University of Science ＆ Technology,Business and Management College,Chongqing 401331, China;
3. Chongqing Rail Transit (Group) Co. Ltd, Chongqing 401120, China;
4. Key Laboratory of Roads and Railway Engineering Safety Control of Ministry of Education,Shijiazhuang Tiedao University,Shijiazhuang, Hebei 050043, China |
|
|
|
Abstract Research purposes: Surface defects of steel rails are important hidden dangers for safe operation of railway traffic, accurate detection of surface defects on steel rails is crucial. The complex environment in which steel rails are in service may cause them to be contaminated with stains, at the same time, the shape of rail defects is often inconsistent. To address the problems of false detection due to stains and difficulty in accurate detection due to different shapes of defects in rail surface defect detection, an improved YOLOv7-based rail surface defect detection method is proposed.
Research conclusions: (1)The problem of false detection of stains was overcome by constructing the dataset with images of rails containing stains as negative samples, with the use of label differences to enable the network to learn features that distinguish between defects and stains. (2)YOLOv7 was improved by deformable convolution with an embedded channel attention mechanism. That was, the deformable convolution replaced the fixed convolution by adding a bias to the convolution sampling points to enhance the network's ability to adapt to the geometric deformation of defects. At the same time, the channel attention mechanism was embedded in the network, and its feature of weighting different channel features made the network focus on the defect features, thus enhancing the defect feature extraction ability. (3) The effectiveness and feasibility of the proposed method was demonstrated by loading the rail surface defect dataset onto the constructed improved YOLOv7 network for end-to-end rail surface defect detection. (4) The research can provide a new method for intelligent detection of rail surface defects.
|
|
Received: 23 April 2022
|
|
|
|
|
|
| [1] |
刘杰, 许建国, 高春丽, 等. 基于机器视觉的接触网吊弦缺陷检测研究[J]. 铁道工程学报, 2022(5): 91-97.Liu Jie, Xu Jianguo, Gao Chunli, etc. Machine Vision - based Research on the Inspection of Dropper Defects of Overhead Contact Line[J]. Journal of Railway Engineering Society, 2022(5): 91-97.
|
| [2] |
张辉, 金侠挺, Wu Q.M. Jonathan, 等. 基于曲率滤波和改进GMM的钢轨缺陷自动视觉检测方法[J]. 仪器仪表学报, 2018(4): 181-194.Zhang Hui, Jin Xiating, Wu Q.M. Jonathan, etc. Automatic Visual Detection Method of Railway Surface Defects Based on Curvature Filtering and Improved GMM[J]. Chinese Journal of Scientific Instrument, 2018(4): 181-194.
|
| [3] |
Song Y N, Zhang H, Liu L, etc. Rail Surface Defect Detection Method Based on YOLOv3 Deep Learning Networks[C]//2018 Chinese Automation Congress (CAC). IEEE, 2018: 1563-1568.
|
| [4] |
Yuan H, Chen H, Liu SH W, etc. A Deep Convolutional Neural Network for Detection of Rail Surface Defect[C]//2019 IEEE Vehicle Power and Propulsion Conference (VPPC). IEEE, 2019: 1-4.
|
| [5] |
Chen X B, Zhang H M.Rail Surface Defects Detection Based on Faster R-CNN[C]//2020 International Conference on Artificial Intelligence and Electromechanical Automation (AIEA). IEEE, 2020: 819-822.
|
| [6] |
罗晖, 徐广隆. 基于图像增强与深度学习的钢轨表面缺陷检测[J]. 铁道科学与工程学报, 2021(3): 623-629.Luo Hui, Xu Guanglong.Rail Surface Defect Detection Based on Image Enhancement and Deep Learning[J]. Journal of Railway Science and Engineering, 2021(3): 623-629.
|
| [7] |
韩强, 刘俊博, 冯其波, 等. 基于多层级特征融合的钢轨表面伤损检测方法[J]. 中国铁道科学, 2021(5): 41-49.Han Qiang, Liu Junbo, Feng Qibo, etc. Damage Detection Method for Rail Surface Based on Multi-Level Feature Fusion[J]. China Railway Science, 2021(5): 41-49.
|
| [8] |
刘俊博, 杜馨瑜, 王胜春, 等. 基于少样本学习的钢轨表面缺陷检测方法[J]. 铁道学报, 2022(7): 72-79.Liu Junbo, Du Xinyu, Wang Shengchun, etc. Rail Surface Defect Detection Method Based on Few-shot Learning[J]. Journal of the China Railway Society, 2022(7):72-79.
|
| [9] |
Li H, Wang F, Liu J, etc. Ensemble Model for Rail Surface Defects Detection[J]. PLoS one, 2022(5): e0268518.
|
| [10] |
Wang C Y, Bochkovskiy A, Liao H Y M. YOLOv7: Trainable Bag-of-freebies Sets New State-of-the-art for Real-time Object Detectors[J]. ArXiv Preprint ArXiv:2207.02696, 2022.
|
| [11] |
Zhu X ZH, Hu H, Lin S, etc. Deformable Convnets v2: More Deformable, Better Results[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019: 9308-9316.
|
| [12] |
Woo S, Park J, Lee J Y, etc. Cbam: Convolutional Block Attention Module[C]//Proceedings of the European Conference on Computer Vision (ECCV). 2018: 3-19.
|
| [13] |
Faghih-Roohi S, Hajizadeh S, Núnez A, etc. Deep Convolutional Neural Networks for Detection of Rail Surface Defects[C]//2016 International Joint Conference on Neural Networks (IJCNN). IEEE, 2016: 2584-2589.
|
| [14] |
梁勇, 邱荣洲, 李志鹏, 等. 基于YOLO v5和多源数据集的水稻主要害虫识别方法[J]. 农业机械学报, 2022(7): 250-258.Liang Yong, Qiu Rongzhou, Li Zhipeng, etc. Identification Method of Major Rice Pests Based on YOLO v5 and Multi-source Datasets[J]. Transactions of the Chinese Society for Agricultural Machinery, 2022(7): 250-258.
|
|
|
|
2024, Vol. 41 
