An Online Rail Track Fastener Classification System Based on YOLO Models
In order to save manpower on rail track inspection, computer vision-based methodologies are developed. We propose utilizing the YOLOv4-Tiny neural network to identify track defects in real time. There are ten defects covering fasteners, rail surfaces, and sleepers from the upward and six defects abo...
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2022-12-01
|
| Series: | Sensors |
| Subjects: | |
| Online Access: | https://www.mdpi.com/1424-8220/22/24/9970 |
| _version_ | 1797455329921859584 |
|---|---|
| author | Chen-Chiung Hsieh Ti-Yun Hsu Wei-Hsin Huang |
| author_facet | Chen-Chiung Hsieh Ti-Yun Hsu Wei-Hsin Huang |
| author_sort | Chen-Chiung Hsieh |
| collection | DOAJ |
| description | In order to save manpower on rail track inspection, computer vision-based methodologies are developed. We propose utilizing the YOLOv4-Tiny neural network to identify track defects in real time. There are ten defects covering fasteners, rail surfaces, and sleepers from the upward and six defects about the rail waist from the sideward. The proposed real-time inspection system includes a high-performance notebook, two sports cameras, and three parallel processes. The hardware is mounted on a flat cart running at 30 km/h. The inspection results about the abnormal track components could be queried by defective type, time, and the rail hectometer stake. In the experiments, data augmentation by a Cycle Generative Adversarial Network (GAN) is used to increase the dataset. The number of images is 3800 on the upward and 967 on the sideward. Five object detection neural network models—YOLOv4, YOLOv4-Tiny, YOLOX-Tiny, SSD512, and SSD300—were tested. The YOLOv4-Tiny model with 150 FPS is selected as the recognition kernel, as it achieved 91.7%, 92%, and 91% for the mAP, precision, and recall of the defective track components from the upward, respectively. The mAP, precision, and recall of the defective track components from the sideward are 99.16%, 96%, and 94%, respectively. |
| first_indexed | 2024-03-09T15:51:53Z |
| format | Article |
| id | doaj.art-bcda8738066e4601b42b3dc4718f2994 |
| institution | Directory Open Access Journal |
| issn | 1424-8220 |
| language | English |
| last_indexed | 2024-03-09T15:51:53Z |
| publishDate | 2022-12-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Sensors |
| spelling | doaj.art-bcda8738066e4601b42b3dc4718f29942023-11-24T17:58:22ZengMDPI AGSensors1424-82202022-12-012224997010.3390/s22249970An Online Rail Track Fastener Classification System Based on YOLO ModelsChen-Chiung Hsieh0Ti-Yun Hsu1Wei-Hsin Huang2Department of Computer Science and Engineering, Tatung University, Taipei 104, TaiwanDepartment of Computer Science and Engineering, Tatung University, Taipei 104, TaiwanThe Graduate Institute of Design Science, Tatung University, Taipei 104, TaiwanIn order to save manpower on rail track inspection, computer vision-based methodologies are developed. We propose utilizing the YOLOv4-Tiny neural network to identify track defects in real time. There are ten defects covering fasteners, rail surfaces, and sleepers from the upward and six defects about the rail waist from the sideward. The proposed real-time inspection system includes a high-performance notebook, two sports cameras, and three parallel processes. The hardware is mounted on a flat cart running at 30 km/h. The inspection results about the abnormal track components could be queried by defective type, time, and the rail hectometer stake. In the experiments, data augmentation by a Cycle Generative Adversarial Network (GAN) is used to increase the dataset. The number of images is 3800 on the upward and 967 on the sideward. Five object detection neural network models—YOLOv4, YOLOv4-Tiny, YOLOX-Tiny, SSD512, and SSD300—were tested. The YOLOv4-Tiny model with 150 FPS is selected as the recognition kernel, as it achieved 91.7%, 92%, and 91% for the mAP, precision, and recall of the defective track components from the upward, respectively. The mAP, precision, and recall of the defective track components from the sideward are 99.16%, 96%, and 94%, respectively.https://www.mdpi.com/1424-8220/22/24/9970railway track inspectiondeep learningobject detection neural networkYOLOreal time |
| spellingShingle | Chen-Chiung Hsieh Ti-Yun Hsu Wei-Hsin Huang An Online Rail Track Fastener Classification System Based on YOLO Models Sensors railway track inspection deep learning object detection neural network YOLO real time |
| title | An Online Rail Track Fastener Classification System Based on YOLO Models |
| title_full | An Online Rail Track Fastener Classification System Based on YOLO Models |
| title_fullStr | An Online Rail Track Fastener Classification System Based on YOLO Models |
| title_full_unstemmed | An Online Rail Track Fastener Classification System Based on YOLO Models |
| title_short | An Online Rail Track Fastener Classification System Based on YOLO Models |
| title_sort | online rail track fastener classification system based on yolo models |
| topic | railway track inspection deep learning object detection neural network YOLO real time |
| url | https://www.mdpi.com/1424-8220/22/24/9970 |
| work_keys_str_mv | AT chenchiunghsieh anonlinerailtrackfastenerclassificationsystembasedonyolomodels AT tiyunhsu anonlinerailtrackfastenerclassificationsystembasedonyolomodels AT weihsinhuang anonlinerailtrackfastenerclassificationsystembasedonyolomodels AT chenchiunghsieh onlinerailtrackfastenerclassificationsystembasedonyolomodels AT tiyunhsu onlinerailtrackfastenerclassificationsystembasedonyolomodels AT weihsinhuang onlinerailtrackfastenerclassificationsystembasedonyolomodels |