Squatflaws, when left unsupervised, can pose a threat to the safety of railway traffic. Unfortunately, as surfaceflaws, they cannot be detected by traditional ultrasonic methods. The application of eddy current to their detection is extremely difficult in the conditions occurring on the railway track. Application of the 2D wavelet transform to the extraction of the rail from the image (the ROI part of the algorithm) allows for fast and reliable detection. The proper selection of wavelet type is a fundamental issue with a significant influence on the performance of the ROI. The algorithm is robust in the varying conditions occurring on the railway track. On the other hand, squat detection is a difficult task, especially when the rail is heavily covered with impurities. Impurities can be drops of oil, dirt, soil or sand covering the rails. In this case, methods based on edge detection fail. It turned out that the approach based on a Gabor filter bank along with a SVM classifier could be useful in devising a solution for this problem. A crucial factor in this method is the correct selection of the number of filters and their parameters. The author decided to use the wrapper SFS method in the selection of salient features. Due to this, the number of features was reduced to six. The SVM with a polynomial kernel of the order of 3 was used as a classifier. The algorithm was implemented in MATLAB and can be run in offline mode. The performance of the system was tested on data not taking part in the process of feature selection and its classification rate was 93%. This promising result allows us to think that it could be an alternative solution to the problem of squat detection.
Visual algorithms for automatic detection of squat flaws in railway rails
Insight ; 55 , 7 ; 353-359
2013
7 Seiten, 11 Bilder, 3 Tabellen, 22 Quellen
Article (Journal)
English
Squat formation on train rails: Growth
Tema Archive | 2012
|Squat formation on train rails: Origination
Tema Archive | 2012
|Studs: a squat-type defect in rails
SAGE Publications | 2012
|Studs: a squat-type defect in rails
Tema Archive | 2012
|