Abstract
This thesis investigates the use of an Optical Time Domain Reflectometry (OTDR) device for railway safety improvement. OTDR sensing, often also termed Distributed Acoustical Sensing (DAS), measures the Rayleigh backscattering of a light pulse along an optical fiber. The resulting signal provides information on local acoustic pressure at linearly spaced segments, corresponding to positions, along the fiber. Using optical time-domain reflectometry, vibrations in the ground caused by different sources can be detected with high accuracy in time and space. We propose a novel method for the detection of vibrations caused by trains in an optical fiber buried within a few meters from the railway track. The presented method learns the characteristic pattern in the Fourier domain using a Support Vector Machine (SVM) and it becomes robust to background noise in the signal. We show that using a General Purpose Graphical Processing Unit (GPGPU) it is possible to compute feature values relevant for train detection in real-time. For the tracking of trains, a point-based causal algorithm is presented. The tracking has two stages to minimize the influence of false classifications of the vibration detection and is solved as an optimization problem. While several algorithms have been demonstrated in the literature for train tracking using OTDR signals, they have neither been tested on longer recordings nor with a large number of train samples. In contrast to that, our data contain four railway stations and more than ten train trajectory crossings over two hours under realistic conditions. To our knowledge, the presented algorithm is the first one in the literature which is tested against ground truth of train trajectories from a conventional train tracking system.
Reference
Papp, A. (2016). Detektion und Verfolgung von Zügen in OTDR Signalen : ein robustes Framework zur Zuglokalisierung in Signalen von optischer Zeitbereichsreflektometrie mit GPGPU Beschleunigung [Diploma Thesis, Technische Universität Wien]. reposiTUm. https://doi.org/10.34726/hss.2016.37305