学术文献库

Millions of human casualties resulting from train accidents globally are caused by the inefficient, manual track inspections. Government agencies are seriously concerned about the safe operations of the rail industry after series of accidents reported across e USA and around the globe, mainly attributed to track defects. Casualties resulting from track defects result in billions of dollars loss in public and private investments and loss of revenue due to downtime, ultimately resulting in loss of the public's confidence. The manual, mundane, and expensive monitoring of rail track safety can be transform through the use of drones, computer vision, and machine learning. The primary goal of this study is to develop multiple algorithms that implement supervised and semi-supervised learning that accurately analyze whether a track is safe or unsafe based on simulated training data of train tracks. This includes being able to develop a Convolutional Neural Network that can identify track defects using supervised learning without having to specify a particular algorithm for detecting those defects, and that the new model would both speed up and improve the quality of the track defect detection process, accompanied with a computer vision image-processing algorithm. Our other goals included designing and building a prototype representation of train tracks to simulate track defects, to precisely and consistently conduct the visual inspection using drones. Ultimately, the goal demonstrates that the state of good repairs in railway tracks can be attained through the use of drones, computer vision and machine learning.