Development of autonomous UAV algorithms for real-time rail tracking and robust deep learning based defect detection

In recent years, the use of rail transport has increased significantly with the development of high-speed trains. As a result of this demand for railway transportation, the length of the railway in the world and in our country is greatly increased. Therefore, the demand for railway maintenance has also increased rapidly and become urgent. Inspection for railway maintenance is an important task to look for and verify defects in railway components. Defects that may occur in railway components significantly affect transportation safety. Control methods currently used depend on contact measurement techniques and human perception. This is why traditional auditing is slow, subjective, and inefficient. In order to overcome these disadvantages and limitations, a new autonomous UAV (Unmanned Aerial Vehicle) computer vision-based method is proposed in this study. The use of Unmanned Aerial Vehicles for the detection of railway defects is a cost-effective approach that offers lower inspection times than traditional techniques. In recent years, UAVs, improvements in processing power, and the shrinking of sensors and their components have increased interest for both military and civilian applications. In particular, the increase in flight times shows that UAVs are at a level that can be used in the field of fault detection. In the thesis, algorithms were developed for autonomous tracking of the railway by the UAV and deep learning-based defect detection was performed on the images obtained by the UAV.