Enhancing Traffic Flow and Driving Safety via Artificial Intelligence

US Transportation Collection2021PDFTech ReportShahmoradi, AmirOsborne, JoshuaSapkota, ParvatBryant, ChristopherDavachi, NiyoushaUniversity of Texas at ArlingtonCenter for Transportation, Equity, Decisions & Dollars (CTEDD). University of TexasUnited States. Department of Transportation. University Transportation Centers (UTC) ProgramUnited States. Department of Transportation. Research and Innovative Technology AdministrationCenter for Transportation, Equity, Decisions and Dollars (CTEDD)TexasUnited StatesArtificial intelligenceCamerasDetection and identification systemsFeasibility analysisMachine visionPrototypesReal time informationTraffic flowTraffic incidentsTraffic engineeringData scienceTraffic CameraAnomaly DetectionCTEDD 019-23USDOT - 69A3551747134Final ReportMotivated by the priorities highlighted by Texas Department of Transportation (TXDOT) and following the guidelines in the recent presidential \u201cExecutive Order on Maintaining American Leadership in Artificial Intelligence\u201d in 2019, this proposal aims to utilize the state-of-the-art tools and techniques in the field of Artificial Intelligence and Data Science to automatically identify and report traffic-related anomalies and hazards using live traffic camera footage across major highways and arterial roads in the State of Texas. Examples of such hazards that are the focus of this proposal include major vehicle-wildlife and vehicle-debris encounters (VWEs and VDEs respectively). Our work builds on top of the existing massive body of literature and research at the intersection of Computer Vision and Traffic Engineering. However, to the extent of our knowledge, this proposal is the first attempt to study the development of an automated pipeline for the detection and reporting of VWEs and VDEs using live traffic camera data. To this aim, we outline and investigate the feasibility of our approach to set up a prototype of a pipeline for real-time collection of data, its reduction, segmentation, analysis, and finally, drawing traffic engineering conclusions and recommendations based on the detected patterns or anomalies in the analysis. This exploratory investigation aims to provide a comprehensive review and proof-of-concept to pave the way towards the next-step proposal and implementation of a commercial-scale version of such data-analytics pipeline in collaboration with the potential major stakeholders, in particular, within Dallas-Fort Worth Metroplex.121