Indexing and matching scheme for recognizing 3D objects from single 2D image

Yoon Sik Tak, Eenjun Hwang

Research output: Chapter in Book/Report/Conference proceedingConference contribution

4 Citations (Scopus)

Abstract

3D object recognition and its pose estimation for a single view query image is useful in the applications such as robot vision, place recognition, medical image analysis and manufacture automation. One straightforward method is to consider all possible view images of objects. Even though the number of view images can be reduced further, still this method requires a huge amount of storage space and computational time. Hence, most of previous works on 3D object recognition and its pose estimation were based on high-end desktop or server platform. However, with common deployment of mobile devices, object recognition and its pose estimation from captured image need to be done in a near real time by the mobile devices. In this paper, we propose a scheme for 3D object recognition and pose estimation on a mobile platform. For object recognition, we utilize the bisymmetric shape property of most objects. For pose estimation, we use SIFT (Scale Invariant Feature Transform) for candidate view images which have same shape features. Also, in order to support such functionality in a mobile platform, we propose a client-server collaboration scheme where the task can be done in a load-balanced way without sacrificing matching accuracy. Through various experiments, we show that our scheme can achieve excellent performance.

Original languageEnglish
Title of host publicationProceedings of the 13th IASTED International Conference on Internet and Multimedia Systems and Applications, IMSA 2009
Pages60-67
Number of pages8
Publication statusPublished - 2009
Event13th IASTED International Conference on Internet and Multimedia Systems and Applications, IMSA 2009 - Honolulu, HI, United States
Duration: 2009 Aug 172009 Aug 19

Publication series

NameProceedings of the 13th IASTED International Conference on Internet and Multimedia Systems and Applications, IMSA 2009

Other

Other13th IASTED International Conference on Internet and Multimedia Systems and Applications, IMSA 2009
CountryUnited States
CityHonolulu, HI
Period09/8/1709/8/19

Keywords

  • 3D object retrieval
  • Distance curve
  • Pose estimation
  • SIFT
  • Shape-based retrieval

ASJC Scopus subject areas

  • Computer Graphics and Computer-Aided Design
  • Computer Networks and Communications
  • Computer Science Applications

Fingerprint Dive into the research topics of 'Indexing and matching scheme for recognizing 3D objects from single 2D image'. Together they form a unique fingerprint.

  • Cite this

    Tak, Y. S., & Hwang, E. (2009). Indexing and matching scheme for recognizing 3D objects from single 2D image. In Proceedings of the 13th IASTED International Conference on Internet and Multimedia Systems and Applications, IMSA 2009 (pp. 60-67). (Proceedings of the 13th IASTED International Conference on Internet and Multimedia Systems and Applications, IMSA 2009).