The cost and effort of modelling existing rail infrastructure from point clouds currently outweigh the perceived benefits of the resulting model. The time required for generating a geometric railway information model is roughly ten times greater than laser scanning it. Hence, there is a persistent need to automate this process. The preliminary step is automatically detecting masts from air-borne LiDAR data, as their position and function is critical to the subsequent detection of other elements. Our method tackles the challenge above by leveraging the highly regulated and standardized nature of railways. It starts with reducing the arbitrary positioning and orientation of the point cloud; and then restricting the search for masts relative to the track centerline. The method verifies the masts’ presence with RANSAC algorithm and delivers detected masts as 3D objects. The method was tested on 18 km railway point cloud and achieves an overall detection rate of 94%.