딥러닝 기반 전차선 동편위 분석 방법 개발

초록

The pantograph-catenary system is essential for electrified railways, with correct installation of contact wires being essential to prevent premature wear. This paper focuses on determining the dynamic stagger of contact wires and detecting irregularities. To achieve this, a deep neural network-based approach, along with several image processing techniques, was employed. The segmentation of contact wires was accomplished using the U-net architecture, followed by post-processing utilizing a thinning algorithm. To track a contact strip, a template matching technique was adopted, and a scan line search was conducted to identify contact points between the contact wires and the contact strip. Subsequently, the contact points were converted into dynamic stagger measurements. The system successfully detected irregularities with an accuracy of 89.6 %, and based on the results, the algorithm shows promise for practical use in the dynamic stagger measurements for the contact wires.

키워드