Dynamic Vergence Using Log-Polar Images

  • Authors:

  • C. Capurro;F. Panerai;G. Sandini

  • Affiliations:

  • LIRA Lab, Department of Communication Computer and Systems Science, University of Genova, Italy. E-mail: nautilus@lira.dist.unige.it, giulio@lira.dist.unige.it;LIRA Lab, Department of Communication Computer and Systems Science, University of Genova, Italy. E-mail: nautilus@lira.dist.unige.it, giulio@lira.dist.unige.it;LIRA Lab, Department of Communication Computer and Systems Science, University of Genova, Italy. E-mail: nautilus@lira.dist.unige.it, giulio@lira.dist.unige.it

  • Venue:
  • International Journal of Computer Vision
  • Year:
  • 1997

Quantified Score

Hi-index 0.00

Visualization

Abstract

Vergence provides robot vision systems with a crucial degree offreedom: it enables fixation of points in visual space at differentdistances from the observer. Vergence control, therefore, affectsthe performance of the stereo system as well as the results of motionestimation and tracking and, as such, must satisfy differentrequirements in order to be able to provide not only a stablefixation, but a stable binocular fusion, and a fast, smooth andaccurate reaction to changes in the environment. To obtain this kindof performance the paper focuses specifically on the use of dynamicvisual information to drive vergence control. In this context,moreover, the use of a space-variant, anthropomorphic sensor isdescribed and some advantages in relation to vergence control arediscussed to demonstrate the relevance of image plane geometry forthis particular task. Expansion or contraction patterns and thetemporal evolution of the degree of fusion measured in the log-polardomain are the inputs to the vergence control system and determinerobust and accurate steering of the two cameras. Real-timeexperiments are presented to demonstrate the performance of thesystem covering different key situations.