Incremental Focus of Attention for Robust Vision-Based Tracking

  • Authors:
  • Kentaro Toyama;Gregory D. Hager

  • Affiliations:
  • Microsoft Research, One Microsoft Way, Redmond, WA 98052-6399. kentoy@microsoft.com;Department of Computer Science, The Johns Hopkins University, Baltimore, Maryland 21218-2694. hager@cs.jhu.edu

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

Quantified Score

Hi-index 0.00

Visualization

Abstract

We present the Incremental Focusof Attention (IFA) architecture for robust, adaptive, real-time motiontracking. IFA systems combine several visual search and vision-based trackingalgorithms into a layered hierarchy. The architecture controlsthe transitions between layers and executes algorithms appropriate tothe visual environment at hand: When conditions are good, tracking isaccurate and precise; as conditions deteriorate, more robust, yet lessaccurate algorithms take over; when tracking is lost altogether, layerscooperate to perform a rapid search for the target and continue tracking.Implemented IFA systems are extremely robust to most common types oftemporary visual disturbances. They resist minor visual perturbancesand recover quickly after full occlusions, illumination changes, majordistractions, and target disappearances. Analysis of the algorithm‘srecovery times are supported by simulation results and experiments onreal data. In particular, examples show that recovery times after losttracking depend primarily on the number of objects visually similar tothe target in the field of view.