Railway Track Faults Detection Using Ensemble Deep Transfer Learning Models

Abstract

Railway track fault detection is an essential task for ensuring the safety and reliability of railway systems, particularly in the summer and rainy seasons when train wheels may slide due to fractures in the track or corrosion may cause track fractures. In this study, we propose a novel approach for the automated detection of railway track faults using deep transfer learning models. The proposed method combines image processing techniques and the training of three pretrained models: InceptionV3, ResNet50V2, and VGG16, on a dataset of railway track images. We evaluated the performance of our proposed method by measuring its accuracy on a test set of railway track images. The individual training accuracies for InceptionV3, ResNet50V2, and VGG16 were 94.30%, 96.79%, and 94.64%, respectively. We then combined these models using an ensemble approach, which achieved an impressive accuracy of 98.57% on the test set. Our results demonstrate the effectiveness of using deep ensemble transfer learning for railway track fault detection. Moreover, our proposed method can be used as a valuable tool for railway track maintenance and monitoring, which can ultimately lead to the improvement of the safety and reliability of railway systems. Our proposed approach for railway track fault detection using ensemble deep transfer learning models shows promising results, indicating that it has great potential for detecting track faults accurately and efficiently. The proposed method can be used in various railway systems worldwide, ultimately leading to improved safety and reliability for passengers and cargo transportation.