The introduction of four-quadrant gate systems that block both the entrance and exit of traffic lanes to railroad grade crossings has resulted in the need for information regarding highway vehicles within the crossing area when these systems are activated by approaching trains. There is also a widespread and growing need for a low-cost crossing surveillance system that could observe motorist behavior at crossings and detect the presence of pedestrians and bicyclists in the crossing area; the raised, lowered, or altered condition of crossing arms; and the functional status of signal crossing lights.
The objectives of this project were to determine the feasibility of using video for real-time detection of the presence of vehicles and trains at railway grade crossings, and to monitor crossings equipped with gates and signal lights to determine whether these devices are functioning properly.
This IDEA project developed and tested software necessary to implement a video-based grade crossing surveillance system using a neural network-based video detection technology. The neural network must be able to accurately interpret the objects that move across a grade crossing as well as the condition and functioning of the crossing warning system components. The system could be used for such functions as providing alarm signals to motorists in extreme danger, messages to maintenance personnel regarding damaged or malfunctioning crossing system components, data for assessing grade crossing risk, and enforcement of grade crossing violations.
The final report for this IDEA project can be found at:
