Automated Video-Based Traffic Count Analysis

In this research, the authors propose a method to extract traffic information from videos of road intersections using a vehicle detection, tracking, and 3D reconstruction system. The video is taken from a low-resolution camera fixed at a relatively low position, about 2 to 7m above ground. First, minimum eigenvalue feature points are detected and tracked with the Kanade-Lucas-Tomasi tracker to obtain feature point movements. These points are then segmented into groups corresponding to vehicles according to features such as position, movement and color. With the information of vehicle's grouping, the 3D position of each feature point can be calculated so that the vehicle's size, accurate position on road, time of entry and exit can be determined. The proposed method is robust to vehicles with varying movement and large perspective deformations.

Language

  • English

Media Info

  • Media Type: Digital/other
  • Edition: Final Research Report
  • Features: Figures; Photos;
  • Pagination: 5p

Subject/Index Terms

Filing Info

  • Accession Number: 01603514
  • Record Type: Publication
  • Contract Numbers: DTRT12GUTG11
  • Files: UTC, NTL, TRIS, RITA, ATRI, USDOT
  • Created Date: May 27 2016 9:52AM