Enhanced bolt positioning for electrified railway catenary systems via small object detection and depth-aware visual servoing

View at publisher Summarise with VIRAT
Authors: Zhiyu Nie, Huanlong Liu, Jingyu Xu, Yuqi Liu
Cited by: 0
References: 34

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Efficient and precise bolt maintenance in electrified railway catenary systems is crucial for ensuring both railway safety and operational efficiency. This study presents an innovative bolt positioning method that integrates an enhanced small-object detection algorithm with depth-aware visual servoing. Building on YOLOv5, we introduce multi-dimensional improvements to develop the CBN-YOLO model, resulting in significant performance enhancements—mAP@0.5 increases by 4.26%, and mAP@0.5:0.95 improves by 4.93%. The method utilizes the four corner points of the bolt’s detection frame as image features for visual servoing, coupled with real-time depth camera data to construct the deep online Jacobian matrix. An image-based visual servo (IBVS) system employing proportional-integral-derivative (PID) control is subsequently designed to achieve accurate bolt localization. Experimental validation on a visual servo test platform demonstrates that the proposed method significantly outperforms the baseline approach of YOLOv5s with fixed-depth visual servoing. The method achieves a 25.56% reduction in the average number of servoing iterations and a 27.57% decrease in the final positioning error, a metric measured as the center deviation. These results demonstrate the effectiveness of our approach in meeting the stringent requirements of bolt maintenance in electrified railway networks and highlight its potential to advance intelligent robotic systems in this field.

Fields of Study & Subjects

Advanced Neural Network Applications Advanced Vision and Imaging Industrial Vision Systems and Defect Detection

Publication details

Journal / source
Proceedings of the Institution of Mechanical Engineers Part C Journal of Mechanical Engineering Science
Volume
240
Issue
13
Pages
4775–4789
Publication date
30 March 2026
Language
EN