Three Staged Railway Track Fault Diagnosis Using Stereo-Imaging, Canny Edge Filter, and Binary Image Classifier

A risky event can result from a com-promise in railway safety. As a result, regular railway track monitoring is required for the rapid diagnosis of rail defects. Due to their minor nature, rail surface problems such as squats are difficult and time-consuming to be diagnosed by techniques such as Axle Based Acceleration (ABA), Fiber Bragg Grating (FBG), and Ultrasonic flaw detection due to their negligible nature. Squats are the catalyst in rail derailment accidents, which account for 70 % of all railway accidents in Pakistan. This study proposes a visual-based squat detection technique that is a three-stage procedure. The three-stage procedure involves stereoscopic image generation, gradient image filter, and logistic regression for detecting the squats on the railway track surface. The accuracy of the developed algorithm is about 97 percent, which is optimal as compared to previous methodologies. The stated procedure was tested at the Kotri railway station under Pakistan Railways' collaboration.