Real-Time Rule-Based Algorithm for Track Detection of Tram-Train

Tram-Train is a form of rail transport that connects urban centers, suburbs, and small and medium-sized cities. It is characterized by the operation of light rail vehicles on the existing railway infrastructure, which runs on two different tracks. In this paper, we propose to detect the pair position of embedded rail and surface rail by fusing classical image analysis methods for Tram-Train track detection. The sliding window algorithm is used to estimate the rail's initial rail point, and the gradient value is used to detect the rail Our algorithm works in a rule-based method by exploiting railway tracks' invariant features, reducing unnecessary computations and improving accuracy. We experiment on a tram test line located in Osong-eup, Chungcheongbuk-do, South Korea. The experiment confirms that the proposed algorithm can detect tracks in real-time.