ABSTRACT
The article deals with the problem of collecting information about railway traffic lights located on the territory of the Russian Federation. To solve this problem, a trained neural network is used that detects railway traffic lights in high-definition images. These images were taken by a camera installed on a train carriage, continuously photographing the area around the track on the right side towards the train. As a result, a detector model of the specified objects was obtained on high-precision images. The paper proposes algorithms for determining the marking and GPS coordinates at a previously detected railway traffic light. The architecture of the information system has been developed, which will have to store complete information about railway traffic lights. The system should be able to quickly search for information and use this data to plan the maintenance of railway traffic lights.
