Vision Rail
The alarming regularity of train-related accidents is a persistent issue, with someone struck by a train every three hours in the U.S., often leading to grave injuries or fatalities. Our product, Vision Rail is a cutting-edge solution to enhance railroad crossing safety: a computer vision-driven detection system designed to monitor crossings in real-time. Strategically placed cameras capture the crossing area, feeding data to our advanced software, which utilizes algorithms to identify both vehicles and pedestrians as trains draw near. Upon detection of any obstacle, the system immediately alerts train engineers, enabling swift action to mitigate potential hazards. This proactive safety measure promises to significantly increase reaction times and reduce the risk of accidents at one of the most critical junctures in rail transport.
Our design involves a static camera watching the railroad crossing and a computer program written in python. The program uses a combination of different computer vision algorithms to detect objects in grade crossings. First, we manually create a mask to exclude the surrounding area of the crossing, leaving only the region where object detection is required. We also take a photo with no object in the grade crossing as the background image. When a train approaches the grade crossing, an image of the crossing is taken, and the background image is subtracted from the image taken, showing the difference between the two images. After this, noise reduction is applied to the image, and a contour is drawn for any objects in the image. Based on the contour seize, the type of the object in the image is identified.
The program was able to detect all objects on the track and can accurately label the object type 96.4% of the time during testing.
In conclusion, Vision Rail can swiftly detect pedestrians and vehicles at railway crossings, alerting train conductors and engineers in real-time to slow down or stop, ultimately leading to decreased railway casualties.