Session: Track 1-5B: Track Safety

Paper Number: 125416

125416 - Railroad Crossing Intrusion Detection Based on Uav-Image and Open-World Object Detection 

Railway crossing incidents, particularly those resulting from unauthorized access and intrusions, have long been a significant safety concern in the rail industry. These incidents frequently lead to severe economic costs and, more critically, pose a risk to human life. Traditional methods of monitoring and securing railway crossings have been challenged by the complexity and dynamic nature of these environments. In many cases, existing detection systems have struggled to accurately identify and effectively respond to intrusions, highlighting a need for more sophisticated and adaptable solutions.

In response to these challenges, our paper introduces the Railway Crossing Intrusion Detection System (RCIDS), a novel approach designed to revolutionize the way railway crossings are monitored and protected. At the heart of RCIDS is the integration of Unmanned Aerial Vehicle (UAV) technology with advanced open-world object detection techniques. This combination aims to set a new standard in railway crossing surveillance and safety, addressing the limitations of previous systems and adapting to the unpredictable nature of real-world scenarios.

UAVs, equipped with high-resolution cameras, are deployed to perform aerial surveys of railway crossings. The aerial vantage point of these drones offers a suite of advantages. Primarily, it allows for expansive coverage of the railway terrain, capturing detailed images that ground-based infrastructure or manual inspections might miss. This broad perspective ensures a more comprehensive monitoring approach, crucial in detecting and responding to various types of intrusions.

The innovation of RCIDS lies not only in the deployment of UAVs but also in the application of bespoke open-world object detection algorithms. Traditional detection systems often operate within a closed-world framework, only able to recognize and respond to predefined categories of objects or threats. In contrast, our open-world object detection algorithms are built to adapt to the unpredictable and variable nature of intrusions at railway crossings. This capability to identify and alert for both known and unknown objects or activities is a critical advancement, offering a level of vigilance and responsiveness that previous systems could not achieve.

RCIDS is more than just a detection tool; it is an integrated safety framework, meticulously designed for comprehensive surveillance and rapid response. Its ability to adapt to new and unforeseen challenges makes it an invaluable asset in the quest for enhanced railway crossing security. As railway networks continue to grow and evolve, the demands for secure and efficient operations also increase. Systems like RCIDS will play a crucial role in meeting these challenges, safeguarding the integrity of railway operations, and protecting the communities that these networks serve. This innovative approach not only contributes to the prevention of accidents and incidents but also supports the broader goal of maintaining public trust and confidence in rail transportation systems.

Presenting Author: Youzhi Tang University of South Carolina

Presenting Author Biography: TBD

Authors:

Youzhi Tang University of South Carolina
Yu Qian University of South Carolina
Dimitris Rizos University of South Carolina
Nikolaos Vitzilaios University of South Carolina