United States Department of Transportation
i
Computer Learning and AI-Based Investigation of Outward Facing Locomotive Videos for Trespassing Events and Behavior
-
2020-09-30
[PDF-800.33 KB]
Details:
-
Creators:
-
Corporate Creators:
-
Corporate Contributors:
-
Subject/TRT Terms:
-
Publication/ Report Number:
-
Resource Type:
-
Geographical Coverage:
-
Edition:Final Report
-
Corporate Publisher:
-
Abstract:This project explored using the analysis of big data through artificial intelligence (AI) and computer learning (CL) to better understand the leading causes for trespassing incidents. The main data set for the analysis is the images from outward facing cameras located in locomotives. Such data are regularly used for analyzing the events in the case of trespasser fatality or serious accident, but the authors are looking to expand their use to systematically identify/analyze all trespasser events. Their approach is to first develop an automatic “trigger” algorithm when human movement is identified in the outward facing locomotive camera and then, in the long-term, use the video data before and after the trigger event to (1) locate every trespasser event visible from the video (within approved limits) and (2) investigate behavioral trends and potential causal factors through application of artificial intelligence, computer learning, and human models on trespasser events. It is unknown how many risky events take place for each incident/casualty. The outward facing video data, combined with proper analytics and technology, offer an opportunity to identify all trespassing events, not only those reported or those leading to casualties, and then use that enlarged understanding toward more systematic analysis of trespasser events, both from spatial and behavioral perspectives. This exploratory portion of the project concentrated on the suitability and quality analysis of the video footage for such analysis.
-
Format:
-
Funding:
-
Collection(s):
-
Main Document Checksum:
-
Download URL:
-
File Type:
