Multi-sensor approach to map the center line of a railway track

Abstract

Geodatasets on railway infrastructure are useful assets in the transportation sector for routing applications and planning construction and maintenance activities. Though some datasets on railway infrastructure exist, they do not cover all places. The current work focuses on automatic digitalisation of the center line of the railway tracks using three sensors: a Global Navigation Satellite System (GNSS) receiver, an Inertial Measurement Unit (IMU) and a Laser Scanner mounted on a vehicle. Three methods that use the data from these sensors were developed: Conditional Selection of GNSS points, Kalman Filter Framework for combining GNSS and IMU data and Extraction of center line of railway tracks from georeferenced Laser Scanner Point Cloud. The center lines were generated from a test dataset by following six different combinations of these three methods. The test dataset consisted of data collected using these three sensors mounted on a measurement platform that moved ten times on a track over a length of 33 km. The results were compared using reference data digitised from the Aerial Images available in the Geodata Portal of Saxony. It was found that the combination of Conditional Selection of GNSS data points and the Kalman Filter Framework gave the best results for the track on which the measurement platform had moved. On an average, 99.91% of the track was digitalised and the mean perpendicular distance of the track from the reference line was 1.49 meters. For the other parallel tracks in the test area, the results from the combination of all the three methods and the combination of the second and third methods gave the best results. On an average, 71.38% of the first parallel tracks and 11.75% of the second parallel tracks were automatically digitalised.