Abstract
A fast and accurate method to detect and locate pantographs is proposed to solve the safety problem of railways caused by pantograph missing or serious deformation. The deep learning based method firstly uses 10,000 images captured by the camera in front of the high-speed rail pantograph as training samples to perform offline training and to generate a CNN prediction model. Then the model and the improved YOLOv2 are used to implement online testing to the pantograph through the image captured by the camera. Test results are finally returned by a program and an early warning of phenomenons of pantograph loss or severe deformation is provided. Results show that the proposed method detects the pantograph in real time with a detection speed of 55 frames per second on a CPU platform while the average accurate rate is 93.1%.
| Translated title of the contribution | A Fast Detection Method for Safety States of Power Receiving Device on High-Speed Rail Based on Deep Learning |
|---|---|
| Original language | Chinese (Traditional) |
| Pages (from-to) | 109-114 |
| Number of pages | 6 |
| Journal | Hsi-An Chiao Tung Ta Hsueh/Journal of Xi'an Jiaotong University |
| Volume | 53 |
| Issue number | 10 |
| DOIs | |
| State | Published - 10 Oct 2019 |