Research Information System
IEEE International Conference on Intelligent Rail Transportation (ICIRT), Birmingham, United Kingdom, 23 - 25 August 2016, pp.270-275
Determining the position and the alignment of the rails ahead of a train is an essential element of camera based driver support systems, which aim to detect the obstacles around the railway. Effective extraction of the rails through a camera faces many challenges that cause poor visibility due to several factors including but not limited to bad weather conditions, inverse illumination, shadows, rust, and clustered rails. Moreover, the process should be completed almost in real time without using any explicit knowledge about the train speed or the camera parameters. This study presents a robust approach that partitions a video frame into four regions, each of which is filtered by two-dimensional Gabor wavelets at different scales. Considering the size and the orientation range of the rails at each partition, responses of the Gabor wavelets, which are tuned to specific frequencies, enable rail edge enhancement together with noise filtering. The proposed approach has been applied to several different scenarios and the results indicate that an effective method is developed.