Railway track detection for train geographic location based on computer vision

Abstract

In a railway infrastructure, train geographic location (e.g., GPS) must be strengthened to adapt to the network topology (i.e., inside or outside the station, straight or curved line, passages through tunnels). Alternative solutions must be proposed to meet this need. Computer vision is one of these disruptive answers to tackle this challenge. Indeed, this technology gives meaning to geographic location by getting closer to human behaviour (i.e., human eye). This paper presents an approach detecting the rails solely by computer vision and the knowledge of certain dimensions of the railway. A case study on rail signalling is also proposed to apply this approach in a safety context.