Real-time detection of rail defects by machine learning

In this study, real-time detection of defects that may occur in rail components will be investigated with machine learning algorithms. This research aims to detect the faults that will occur in the railway components early and to prevent accidents and losses in advance. The study was carried out by detecting the faults of basic parts such as rail fractures, lack of fastener, lack of rail fixing clamp, and lack of screws. Multiple images will be used in our application for training machine learning algorithms. Defective or non-defective images will be classified by a deep learning network. In other words, our deep learning network will detect incoming test images contactless and in real-time. In our study, appropriate methods and success metrics are given and the results are explained comparatively.