An obstacle detection system for automatic trains

Automatic trains play an important role to increase the traffic share of railways. One necessary function of an automatic system is the detection of obstacles in front of a train. Obstacles can be detected by evaluation of sensor data perceiving the environment. If a train is equipped with a video camera, machine vision approaches can provide knowledge about the actual environment in front of the train. Video sensors, especially in a multisensor configuration, are very useful to record the environment of a moving train due to their high resolution compared to their size and cost. Based on experiences gathered with two experimental vehicles, which have been designed and commissioned to investigate approaches for driver assistance on public roads, we present a system architecture for an automatic train. Our current system architecture for the overall procedure consists of the following components: Man-machine-interaction, route planning, situation modelling, sensor evaluation functions, and vehicle control and driver warning. Experimental results of rail-tracking and obstacle detection by evaluation of image sequences will be presented.