With a growing density of traffic in cities, collision assessment systems for public transportation are emerging and video analysis is broadly accepted as a cornerstone for this application. For trams, the finding of tram tracks is an essential part of such a system to estimate a safety margin of crossing traffic participants. Tram-track detection is a challenging task due to the urban environment in which trams operate, e.g. clutter, sharp curves and occlusions of the track. Existing methods, such as rail-road detection for trains, are not suited for such challenging conditions. In this paper, we present a novel and generic system to detect the tram track in advance of the tram position. The system incorporates inverse perspective mapping and a-priori geometry knowledge of the rails to find possible track segments. We contribute to the state of the art by creating a new track reconstruction algorithm which relies on graph theory. We define track segments as vertices in a graph, in which edges represent feasible connections. This graph is then converted to a max-cost arborescence graph, and the best path is selected according to its location and additional temporal information. Our system outperforms a state-of-the-art traintrack detector. Furthermore, the system performance is validated on 3,600 manually annotated frames. Results are promising, where straight tracks are found in more than 90% of the images, while complete curves are still detected in 37% of the cases.
|Date of Award
|28 Feb 2014
|Sveta Zinger (Supervisor 1), Peter H.N. de With (Supervisor 2), D.W.J.M. van de Wouw (External coach) & Egbert G.T. Jaspers (External coach)