A novel integrated multi-sensor on-board obstacle detection system for freight trains, developed within the Shift2Rail initiative, combines different vision technologies to identify obstacles beyond 200m, the current state-of-art, and up to 1000m.
This innovative device, able to operate under any light and weather conditions, consists of a thermal camera, a night vision sensor (camera augmented with image intensifier), a multi stereo-vision system and a laser scanner.
Sensors are easy to mount and dismount from the device housing, enabling different evaluation tests in static and moving vehicles.
A computer vision software, based on machine learning, supports the hardware in analysing images at high speed, determining distance, shapes and sizes of obstacles.
A major trial was conducted with a 21-wagon freight train, pulled by a 444 ŽS series locomotive, running for 120 km on the Serbian section of the Pan European corridor X in July 2018. The device showed reliable obstacle detection up to 500m.
Current results are a good basis for further advancing the software to achieve reliable long-range (up to 1,000 m) autonomous detection of obstacles on the rail tracks.
This device is an important contribution to the development of autonomous rail freight systems, necessary for the implementation of the EU transport strategy.