Detection of Safe Passage for Trains at Rail Level Crossings Using Deep Learning.

Area of Science:

• Railway Engineering
• Computer Vision
• Artificial Intelligence

Background:

• Ensuring train traffic safety at rail level crossings (RLC) is critical.
• Reliable sensors are essential for traffic control systems to monitor RLC status.
• Information fusion from multiple sensors improves situational awareness and reaction capabilities.

Purpose of the Study:

• To develop and validate a deep learning method for processing video data to determine the operational state of RLCs.
• To enhance the safety of train passage by accurately assessing the region of interest (ROI) at RLCs.
• To create a robust system for obstacle detection at RLCs.

Main Methods:

• Utilized deep learning algorithms for video data processing.
• Implemented a method to determine the state of the region of interest (ROI) vital for safe train passage.
• Validated the approach using extensive video surveillance data from various RLC sites.

Main Results:

• Achieved high recall values of 0.98.
• Demonstrated effective obstacle detection across diverse weather conditions and seasons.
• Showcased significantly reduced processing resource requirements.

Conclusions:

• The proposed deep learning method offers a reliable solution for obstacle detection at RLCs.
• The system can serve as an auxiliary signal source for train control systems.
• Fused data from this method and other sensors can meet stringent railway safety standards.