Refinements and evaluations of line-based pose enumeration from a single image

Takeshi Shakunaga

Research output: Contribution to journalArticle

Abstract

This paper proposes robust algorithms for linebased pose enumeration from a single view, and it reports on their evaluations by simulations. The proposed algorithms incorporate two major refinements into the algorithms originally proposed by Shakunaga[I]. The first refinement, introduction of zone-crossing detection to the I-d search remarkably decreases the rate of overlooking a correct pose. The second refinement, adaptive selection of a PAT pair considerably reduces the average estimation error. Simulation results show that pose estimation precision depends primarily on the precision of line detection. Although the refinements are widely effective, they are more effective for more precise line detection. For 99% of rigid body samples, the algorithm can estimate rotation with an error of less than 2 degrees, and for 99.9% of the samples, the error is less than 10 degrees. Simulation experiments for articulated objects show similar results by using the second algorithm. The effectiveness of the algorithms is verified in an alignment approach by simulations.

Original languageEnglish
Pages (from-to)1266-1273
Number of pages8
JournalIEICE Transactions on Information and Systems
VolumeE79-D
Issue number9
Publication statusPublished - Jan 1 1996

Keywords

  • Evaluation
  • Jointed bodies
  • Line segment
  • Pose enumeration
  • Pose estimation
  • Rigid body

ASJC Scopus subject areas

  • Software
  • Hardware and Architecture
  • Computer Vision and Pattern Recognition
  • Electrical and Electronic Engineering
  • Artificial Intelligence

Fingerprint Dive into the research topics of 'Refinements and evaluations of line-based pose enumeration from a single image'. Together they form a unique fingerprint.

  • Cite this