Li, Xinman (Beijing Jiaotong University), Cao, Yuan (Beijing Jiaotong University), Wang, Feng (Beijing Jiaotong University), Sun, Yongkui (Beijing Jiaotong University), Su, Shuai (Beijing Jiaotong university), Yang, Weifeng (the CRRC Zhuzhou Institute Co, Ltd), wang, Wenkun (CRRC Zhuzhou Institute Co, Ltd)

Rail Fastener Defect Detection of Heavy Haul Railway Based on Improved YOLOv8

Scheduled for presentation during the Regular Session "Sensing, Vision, and Perception II" (WeBT2), Wednesday, September 25, 2024, 16:10−16:30, Salon 5

2024 IEEE 27th International Conference on Intelligent Transportation Systems (ITSC), September 24- 27, 2024, Edmonton, Canada

This information is tentative and subject to change. Compiled on December 26, 2024

Abstract

Rail fasteners play a crucial role in the heavy haul railway lines, serving as vital components of railway infrastructure. However, due to the intricate nature of the heavy haul railway environment, fasteners can often be obscured, leading to misjudgments and inaccuracies for the existing defect detection algorithms. To address this challenge, a novel method to detect rail fastener defects based on an improved YOLOv8 model is proposed. Initially, a new dataset specifically tailored for fastener defect detection in heavy haul railway lines is constructed utilizing on-site data. To enrich feature information and enhance detection accuracy, multiple improvements are introduced, including the addition of a P6 feature layer, the design of a cross-layer connection neck (CCNeck) feature fusion network, and the introduction of the DySample up-sampling operator. Furthermore, the Quality Focal Loss (QFL) function is introduced to address class imbalance problem within the datasets. Finally, experiments are conducted and the results demonstrate significant enhancements in detection mean average precision (mAP) compared to existing state-of-the-art networks.