A Comparison of Feature Detectors and Descriptors in RGB-D SLAM Methods

GÜÇLÜ O., CAN A. B.

12th International Conference on Image Analysis and Recognition (ICIAR), Niagara Falls, Canada, 22 - 24 July 2015, vol.9164, pp.297-305 identifier identifier

  • Publication Type: Conference Paper / Full Text
  • Volume: 9164
  • Doi Number: 10.1007/978-3-319-20801-5_32
  • City: Niagara Falls
  • Country: Canada
  • Page Numbers: pp.297-305
  • Hacettepe University Affiliated: Yes

Abstract

In RGB-D based SLAM methods, robot motion is generally computed by detecting and matching feature points in image frames obtained from an RGB-D sensor. Thus, feature detectors and descriptors used in a SLAM method significantly affect the performance. In this work, impacts of feature detectors and descriptors on the performance of an RGB-D based SLAM method are studied. SIFT, SURF, BRISK, ORB, FAST, GFTT, STAR feature detectors and SIFT, SURF, BRISK, ORB, BRIEF, FREAK feature descriptors are evaluated in terms of accuracy and speed.