Research on An Intelligent Recognition Method for The Components of the Railway Overhead Line

The railway overhead line system is used to provide a power source for high-speed trains. The importance of the components in the overhead line system is apparent. This paper conducts research on the measurement of key components, such as the contact line, the catenary cable, and the suspension string, and identifies them using a neural network. An auto-encoder is used to reduce the measurement noise and extract characteristics in the image domain. Subsequently, deep neural networks (DNNs) are employed to clarify the pixels in the measurement data to determine the positions and distributions of the components. Testing results indicate that the position errors are less than 3 mm, and the categorization of the components matches the relative objects with $100.0 \%$ accuracy, demonstrating the efficiency and accuracy of the method.