This article describes visual functions dedicated to the extraction and recognition of visual landmarks, here planar quadrangles detected by a single camera. Landmarks are extracted among edge segments through a relaxation scheme, used to apply geometrical, topological and appearance constraints on sets of segments. Once extracted, such a landmark is characterized by invariant attributes so that recognition is made possible from a large range of viewpoints. Landmarks are represented by an icon which is built using the homography between the current viewpoint and a reference shape (a square). When detected again, the landmark is recognized by using a distance between icons. We propose a comparison of several of these metrics and an evaluation on actual and synthetic images that shows the validity of our approach. Results issued from experiments of a mobile robot navigating in an indoor environment are finally presented.
A visual landmark framework for mobile robot navigation
2007-01-01
Aufsatz (Zeitschrift)
Elektronische Ressource
Englisch
DDC: | 629 |
A visual landmark framework for mobile robot navigation
British Library Online Contents | 2007
|Landmark-Oriented Visual Navigation of an Autonomous Mobile Robot in Indoor Environments
British Library Conference Proceedings | 1993
|Landmark robust selection for asteroid landing visual navigation
Elsevier | 2023
|Unidentified landmark navigation.
NTRS | 1967
|