United States Department of Transportation
i
Development of Railroad Trespassing Database Using Artificial Intelligence
-
2024-02-01
-
[PDF-8.00 MB]
Details:
-
Creators:
-
Corporate Creators:
-
Corporate Contributors:
-
Subject/TRT Terms:
-
Publication/ Report Number:
-
Resource Type:
-
Geographical Coverage:
-
Edition:May 2020 – September 2023
-
Contracting Officer:
-
Corporate Publisher:
-
Abstract:The Federal Railroad Administration (FRA) sponsored a research team from Rutgers University to develop a proof-of-concept Trespassing Database using Artificial Intelligence (AI) technology to automatically process large volumes of live or recorded video data. The team used the Rutgers AI algorithm to analyze over 27,000 hours of live video data and 1,176 hours of recorded video data from rights-of-way and grade crossings at 11 locations in 6 states. The AI algorithm collected trespassing-related data, including traffic, rail signal activations, train events, and trespass events. Trespass event data were automatically collected for each trespasser, including date, time, type (e.g., person, car, truck, bus, motorcycle), weather, trespasser’s path, and a video clip. The team manually validated all trespass event detection results to ensure that accurate data was included in the database. Over 29,000 trespass events were detected by the AI algorithm across all studied locations in this research.
-
Format:
-
Funding:
-
Collection(s):
-
Main Document Checksum:
-
Download URL:
-
File Type:
