EXTRACTING VEHICLE TRAJECTORIES FROM VIDEO DATA: THE NGSIM PROTOTYPE DATA SET

The Next Generation Simulation (NGISM) program, initiated by the Federal Highway Administration (FHWA), has as its objective the development of behavioral algorithms that support microscopic traffic simulation, accompanied by documentation and validation sets. This article relates how algorithms from the California Partners for Advanced Transit and Highways (PATH) program were selected to produce a prototype data set of vehicle trajectories as well as aid in assessing the viability of emerging technologies for automating the trajectory generation process. Vehicle trajectories are needed to validate microscopic simulation algorithms. The article describes the development of a new approach in which vehicles are first detected based on their appearance, and then tracked by a separate tracking algorithm. This new appearance-based vehicle detection algorithm extracts line segments from images and matches them to 3D vehicle models. Different cues are used to find the match of the lines segments to the vehicle models. Once a vehicle is detected, it can then be tracked according to its appearance in the camera image. The article relates how a complete set of trajectories from video data were generated through an interactive vehicle tracking system based on the algorithm. Cost testing has shown that the resulting system, because of its significantly smaller computation process, yielded a 93% cost reduction compared to the manual system.

Language

  • English

Media Info

  • Features: Figures; Photos; References;
  • Pagination: 2p
  • Serial:
    • Intellimotion
    • Volume: 11
    • Issue Number: 3
    • Publisher: University of California, Berkeley
    • ISSN: 1061-4311

Subject/Index Terms

Filing Info

  • Accession Number: 01005594
  • Record Type: Publication
  • Source Agency: UC Berkeley Transportation Library
  • Files: BTRIS, TRIS
  • Created Date: Oct 19 2005 2:04PM