RAILWAY TRACK FAULT DETECTION USING COMPUTER VISION

A computer vision technique is used for automatic detection of railway tracks. Interior orientation parameters are extractedas part of the camera calibration task. Bundle adjustment calibration technique is used to compute the exterior orientationparameters based on selected ground control points. Hence, the eye fish effect of the images is removed and orthogonalimages for the railway tracks are constructed using matching and feature extraction algorithms. The framework is testedon a data set of a railway network with an approximate length of 20 km. Accidents due to defective railway lines andderailments are common disasters that are observed frequently in Southeast Asian countries. It is imperative to run properdiagnosis over the detection of such faults to prevent such accidents. However, manual detection of such faults periodicallycan be both time-consuming and costly. In this paper, we have proposed a Deep Learning (DL)-based algorithm forautomatic fault detection in railway tracks, which we termed an Ensembled Convolutional Autoencoder Resnet-basedRecurrent Neural Network (ECARRNET). We compared its output with existing DL techniques in the form of several pretrained DL models to investigate railway tracks and determine whether they are defective or not while consideringcommonly prevalent faults such as—defects in rails and fasteners. Moreover, we manually collected the images fromdifferent railway tracks situated in Bangladesh and made our dataset. After comparing our proposed model with the existingmodels, we found that our proposed architecture has produced the highest accuracy among all the previously existing stateof-the-art (SOTA) architecture, with an accuracy of 93.28% on the full dataset.The accuracy of the results is compared with a field survey data conducted to the same area usingconventional surveying instruments such as Total station and Global Navigation Satellite System(GNSS). The proposed framework enables automatic extraction of railway tracks and its relationship with surroundingfeatures, which contributes to quality control and assurance procedures for field collected data. The framework also offersa method for continuous and low-cost monitoring of the railway network