Hough$^2$Map – Iterative Event-Based Hough Transform for High-Speed Railway Mapping

To cope with the growing demand for transportation on railway system, accurate, robust, and high-frequency positioning is required to enable a safe efficient utilization of existing infrastructure. As basis localization system we propose complete on-board mapping pipeline able map robust meaningful landmarks, such as poles from power lines, in vicinity vehicle. Such are good candidates reliable long term landmarks even through difficult weather conditions or seasonal changes. address challenges motion blur illumination changes scenarios employ Dynamic Vision Sensor, novel event-based camera. Using sideways oriented camera, appear vertical lines. lines real-time event stream, introduce Hough2Map, consecutive iterative Hough transform framework capable detecting, tracking, triangulating close-by structures. We demonstrate reliability accuracy Hough2Map real-world data typical usage evaluate using surveyed infrastructure ground truth maps. achieves detection up 92% root mean square error 1.1518m.