Rail Vision has received a patent from the European Patent Office for a railway collision avoidance method and system that uses AI and electro-optical imaging.

The company said that the protection applies in Europe and relates to a system designed to detect hazards on and near the tracks ahead of a train.

The patented method uses forward-looking single-spectrum or multispectral electro-optical imaging in combination with a specific system architecture and scene analysis based on deep learning.

The technology operates through a Convolutional Neural Network (CNN) that first determines the path of the railway in front of the locomotive or train.

A second object detection CNN then examines the area around that path to identify the possible obstacles in real time.

According to Rail Vision, the system produces alarms for various hazards, including railway switch occurrences and states, obstructions, impending end-of-rail effects and different types of obstacles.

The company noted that the technology supports decision-making for locomotive drivers during manned operations and can also enable automated decision-making for driverless trains.

Rail Vision said that the European patent forms part of its global intellectual property protection strategy and follows earlier patent approvals in the US, Japan, and India.

In August this year, Rail Vision received a Decision to Grant from the European Patent Office for its patent application covering a system that improves the sampling rate of an imager detector for a Selected Region of Interest (SROI).

The patented technology includes an imaging device and a processing unit that work together to capture, analyse, and prioritise visual data from a forward-facing camera mounted on a train or locomotive.