Combination of Edge Element and Optical Flow Estimates for 3D-Model-Based Vehicle Tracking in Traffic Image Sequences

  • Authors:

  • Michael Haag;Hans-Hellmut Nagel

  • Affiliations:

  • Institut für Algorithmen und Kognitive Systeme, Universität Karlsruhe (TH), Postfach 6980, D-76128 Karlsruhe, Germany. haag@computer.org;Fraunhofer-Institut für Informations- und Datenverarbeitung IITB, Fraunhoferstraße 1, D-76131 Karlsruhe, Germany. nagel@ira.uka.de

  • Venue:
  • International Journal of Computer Vision
  • Year:
  • 1999

Quantified Score

Hi-index 0.00

Visualization

Abstract

A model-based vehicle tracking system for the evaluation ofinner-city traffic video sequences has been systematically tested onabout 15 minutes of real world video data. Methodologicalimprovements during preparatory test phases affected—among otherchanges—the combination of edge element and optical flow estimatesin the measurement process and a more consequent exploitation ofbackground knowledge. The explication of this knowledge in the formof models facilitates the evaluation of video data for differentscenes by exchanging the scene-dependent models. An extensive seriesof experiments with a large test sample demonstrates that the currentversion of our system appears to have reached a relative optimum: further interactive tuning of tracking parameters does nolonger promise to improve the overall system performancesignificantly. Even the incorporation of further knowledge regardingvehicle and scene geometry or illumination has to cope with anincreasing level of interaction between different knowledge sourcesand system parameters. Our results indicate that model-based trackingof rigid objects in monocular image sequences may have to bereappraised more thoroughly than anticipated during the recent past.