Citation

Lee H, Hong J, Wendimagegn TW, Lee H. Sensors (Basel) 2021; 21(24): e8335.

Copyright

(Copyright © 2021, MDPI: Multidisciplinary Digital Publishing Institute)

DOI

10.3390/s21248335

PMID

34960429

Abstract

Rail corrugation appears as oscillatory wear on the rail surface caused by the interaction between the train wheels and the railway. Corrugation shortens railway service life and forces early rail replacement. Consequently, service can be suspended for days during rail replacement, adversely affecting an important means of transportation. We propose an inspection method for rail corrugation using computer vision through an algorithm based on feature descriptors to automatically distinguish corrugated from normal surfaces. We extract seven features and concatenate them to form a feature vector obtained from a railway image. The feature vector is then used to build support vector machine. Data were collected from seven different tracks as video streams acquired at 30 fps. The trained support vector machine was used to predict test frames of rails as being either corrugated or normal. The proposed method achieved a high performance, with 97.11% accuracy, 95.52% precision, and 97.97% recall. Experimental results show that our method is more effective in identifying corrugated images than reference state-of the art works.

Language: en

Keywords

cepstrum transformation; frequency-domain analysis; rail corrugation detection; static harmonic feature; support vector machine