Rail Vision has been granted a patent by the European Patent Office for an AI-based collision avoidance method and system in rail transport. The solution combines forward-facing, single- or multi-spectral electro-optical imaging with a proprietary system architecture and deep learning methods for environmental assessment. In a two-stage CNN process, the system first precisely determines the track line being travelled on and then analyses the track area and its surroundings for obstacles or dangerous situations. It generates real-time warnings about the position and condition of switches, obstacles on the track, end-of-rail effects and other collision-relevant objects. The technology serves as a decision-making aid for train drivers and can also support fully automated decisions in driverless operation. The European patent is part of Rail Vision's global IP strategy, following previously granted property rights in the USA, Japan and India.
