Skip to main navigation Skip to search Skip to main content

SIFT in GMS-selected areas: A method for feature point matching between photographs and rendered images

  • Beihang University

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Pose estimation with rendered key images often enables more effective and efficient performance for on-line tasks. However, it is difficult to establish correspondences between photographs and rendered images because of their different sources and features. This paper proposes SIGMA (SIFT in GMS-selected Areas), a method combining high quality detector and descriptor with effective method of removing mismatches. It preforms SIFT locally in the areas selected by GMS, which provides high precision and good distribution of points. Experiments show the precision of SIGMA is 0.7638, higher than SIFT(0.5037) and GMS(0.7206).

Original languageEnglish
Title of host publicationSixth Symposium on Novel Optoelectronic Detection Technology and Applications
EditorsJunhao Chu, Huilin Jiang
PublisherSPIE
ISBN (Electronic)9781510637047
DOIs
StatePublished - 2020
Event6th Symposium on Novel Optoelectronic Detection Technology and Applications - Beijing, China
Duration: 3 Dec 20195 Dec 2019

Publication series

NameProceedings of SPIE - The International Society for Optical Engineering
Volume11455
ISSN (Print)0277-786X
ISSN (Electronic)1996-756X

Conference

Conference6th Symposium on Novel Optoelectronic Detection Technology and Applications
Country/TerritoryChina
CityBeijing
Period3/12/195/12/19

Keywords

  • Computer vision
  • Feature point matching
  • GMS
  • Image processing
  • SIFT

Cite this