6 DOF Pose Estimation for Efficient Robot Manipulation

Hsien I. Lin, Subhajit Nanda

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

1 Scopus citations

Abstract

To manipulate 3D objects, pose estimation techniques are necessary. So, in this paper we proposed a 6D pose estimation algorithm for robotic manipulation, which has several advantages over existing methods: first, it does not need any large data-set and costly computing techniques; and second, it can work with various camera sensor modules. In our proposed method, we integrated two feature matching algorithms for matching the features between the current image and database images to locate the target. RANdom SAmple Consensus (RANSAC) algorithm was used to refine the inlier points of the object from the outliers points. Later, Perspective-n-point (P-n-P) algorithm was adopted to compute the pose of the object. Experiments revealed that the proposed method can locate and compute object poses successfully.

Original languageEnglish
Title of host publicationProceedings - 2020 IEEE Conference on Industrial Cyberphysical Systems, ICPS 2020
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages279-284
Number of pages6
ISBN (Electronic)9781728163895
DOIs
StatePublished - 10 Jun 2020
Event3rd IEEE Conference on Industrial Cyberphysical Systems, ICPS 2020 - Virtual, Tampere, Finland
Duration: 10 Jun 202012 Jun 2020

Publication series

NameProceedings - 2020 IEEE Conference on Industrial Cyberphysical Systems, ICPS 2020

Conference

Conference3rd IEEE Conference on Industrial Cyberphysical Systems, ICPS 2020
Country/TerritoryFinland
CityVirtual, Tampere
Period10/06/2012/06/20

Keywords

  • feature matching
  • Object detection
  • Perspective-n-point
  • pose estimation
  • RANSAC

Fingerprint

Dive into the research topics of '6 DOF Pose Estimation for Efficient Robot Manipulation'. Together they form a unique fingerprint.

Cite this