|
|
|
| High Robust Track Surface Defect Detection Method Based on Computer Vision Technique |
| LI Dong1, WANG Rui1, WANG Ye2, JIANG Zhouxian1, CUI Xiaotong1 |
1. Beijing Jiaotong University, Beijing 100044, China;
2. China Railway Liuyuan Group Co. Ltd, Tianjin 300308, China |
|
|
|
Abstract Research purposes: In order to solve the uncertainty problem of decision-making caused by the decrease of robustness of the track defect detection system under the interference of actual natural environmental factors, a method to improve the robustness of the detection system is proposed. Firstly, the mutation algorithm is used to generate diversified robust training samples, so that the model can learn to adapt to various environmental disturbances and changes. Then, the target detection model based on YOLOv5 is selected as the track defect detector to meet the requirements of high precision, real-time performance and high accuracy.
Research conclusions: (1)Through the mutation generation algorithm, more diverse robust training samples can be generated, so that the model has more samples to learn various disturbances and changes, so as to adapt to various environments. (2)The selected YOLOv5 target detection model as a track defect detector can not only meet the high requirements of track defect detection for detection accuracy and real-time performance, but also achieve good detection results with its excellent generalization ability even if the number of training samples is limited. It fits the difficulty of data acquisition in the actual track detection scene and shows strong applicability. (3) Experiments were carried out on the track inspection data collected on real subway lines. The experimental results show that the robustness of the robust retrained track defect detector on the disturbance data has been significantly improved. The average accuracy rate is increased by 23.35 %, the recall rate is increased by 32.75 %, the mAP50 is increased by 30.98 %, and the mAP50-95 is increased by 19.54 %. (4)The research results can be applied to railway, subway and other transportation fields, which is helpful to ensure the safe operation of the line.
|
|
Received: 28 September 2023
|
|
|
|
|
|
| [1] |
雷冬,杜文康,朱国靖,等.基于机器视觉方法的高铁桥梁监测技术研究[J].铁道工程学报,2023(3):45-49.Lei Dong, Du Wenkang, Zhu Guojing, etc. Research on the Monitoring Technology of High-speed Railway Bridge Based on Machine Vision Method[J]. Journal of Railway Engineering Society, 2023(3):45-49.
|
| [2] |
Song H, Sun Z, Wang H, etc. Enhancing Train Position Perception through AI-driven Multi-source Information Fusion[J]. Control Theory and Technology, 2023(3): 425-436.
|
| [3] |
Song H, Gao S, Li Y, etc. Train-centric Communication Based Autonomous Train Control System[J]. IEEE Transactions on Intelligent Vehicles, 2022(1): 721-731.
|
| [4] |
Redmon J, Divvala S, Girshick R, etc. You Only Look Once: Unified, Real-time Object Detection[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition,2016: 779-788.
|
| [5] |
Mandriota C, Nitti M, Ancona N, etc. Filter-based Feature Selection for Rail Defect Detection[J]. Machine Vision and Applications, 2004, 15: 179-185.
|
| [6] |
闵永智,岳彪,马宏锋,等.基于图像灰度梯度特征的钢轨表面缺陷检测[J].仪器仪表学报,2018(4):220-229.Min Yongzhi, Yue Biao, Ma Hongfeng, etc. Rail Surface Defects Detection Based on Gray Scale Gradient Characteristics of Image[J]. Chinese Journal of Scientific Instrument, 2018(4):220-229.
|
| [7] |
Gibert X, Patel V M, Chellappa R.Robust Fastener Detection for Autonomous Visual Railway Track Inspection[C]//2015 IEEE Winter Conference on Applications of Computer Vision. IEEE, 2015:694-701.
|
| [8] |
Zhang X, Wang K, Wang Y, etc. An Improved Method of Rail Health Monitoring Based on CNN and Multiple Acoustic Emission Events[C]//2017 IEEE International Instrumentation and Measurement Technology Conference (I2MTC). IEEE, 2017:1-6.
|
|
|
|
2024, Vol. 41 
