AUTOMATIC TRAFFIC MONITORING FROM VIDEO IMAGES . 15TH ARRB CONFERENCE, DARWIN, NORTHERN TERRITORY, 26-31 AUGUST 1990; PROCEEDINGS PARTS 1 TO 7

This paper describes a novel approach to the computer analysis of the video images returned by roadside surveillance cameras. The aim is to provide a quantitative description of the state of the traffic within the camera's field of view. The research presented addresses the problem of describing traffic patterns directly from the image rather than via measurement of individual vehicle parameters such as speed and location. The primary objective has been to design an algorithm which enables the continuous monitoring of traffic surveillance cameras to be performed automatically. An operator's attention can then be drawn to the appropriate camera when the occurrence of a pre-determined event, such as a queue or an incident, is detected. An outline is given of the need for traffic surveillance and some of the current methods used are surveyed. The utility of image processing for monitoring is discussed and the strategy adopted in this research described. A description of how the spatial pattern of traffic at an instant in time can be generated from images transformed by these methods is given. An evaluation method for the algorithm is described together with the results for trial runs on an example scene and discussion of the sources of error. Finally, the performance of the algorithm is considered in the light of the test results. (Author/TRRL)

  • Availability:
  • Corporate Authors:

    ARRB

    Melbourne, Victoria  Australia 
  • Authors:
    • HOOSE, N
  • Publication Date: 1990

Language

  • English

Media Info

Subject/Index Terms

Filing Info

  • Accession Number: 00609370
  • Record Type: Publication
  • Source Agency: Transport and Road Research Laboratory (TRRL)
  • Files: ITRD, TRIS
  • Created Date: Jun 30 1991 12:00AM