Model-based rail detection in mobile laser scanning data

摘要: Similar to autonomous vehicles, future train applications require an accurate on-board self-localization for railway vehicles. Therefore, a reliable and real-time capable environment perception is required. In particular, the knowledge of track taken at turnout overcomes ambiguities in self-localization. As most important groundwork this, paper introduces new approach detection rails tracks solely from 2d lidar measurements. The technique based on feature point method data, template matching approach, spatial clustering extract detected rail elements. evaluated six different datasets outdoors demanding test ground. It provides results with centimeter accuracy, recall about 90 %, precision 95 %. able detect even complex real-world topologies such as turnouts more than two rails.