Automatic detection of rail track position from images captured from a train mounted video camera has many useful applications. The aim of the work presented in this paper is to develop a system for such rail track detection, using image processing techniques. The overlapping, merging and splitting of rails presents a source of difficulty for any rail detection method. Hence, we present a novel method which is capable of extracting the true rails in an image, despite their interactions, by way of pairing left and right hand rails into complete tracks. This pairing is achieved using a region growing-based method, utilizing spatial constraints and candidate rail track positions to group together semantically correct image segments into tracks. Paired rails are then efficiently stored by fitting parametric curves to candidate rail track data, allowing only a few control points to be stored per rail. Results show that the methods developed provide a robust and compact representation of rails present in test images.
A novel rail pairing method for compact rail track representation
Eine neuartige, kompakte Methode zur Erkennung der Parallelität von Schienensträngen
2010
6 Seiten, 9 Bilder, 11 Quellen
(nicht paginiert)
Aufsatz (Konferenz)
Datenträger
Englisch
A novel rail pairing method for compact rail track representation
Kraftfahrwesen | 2010
|REMOTE RAIL TRACK LAYING DEVICE AND REMOTE RAIL TRACK LAYING METHOD
Europäisches Patentamt | 2016
|