Title :
An active strategy for plane detection and estimation with a monocular camera
Author :
Giordano, Paolo Robuffo ; Spica, Riccardo ; Chaumette, Francois
Author_Institution :
Irisa & Inria Rennes Bretagne Atlantique, Rennes, France
fDate :
May 31 2014-June 7 2014
Abstract :
Plane detection and estimation from visual data is a classical problem in robotic vision. In this work we propose a novel active strategy in which a monocular camera tries to determine whether a set of observed point features belongs to a common plane, and, if so, what are the associated plane parameters. The active component of the strategy imposes an optimized camera motion (as a function of the observed scene) able to maximize the convergence in estimating the scene structure. Based on this strategy, two methods are then proposed to solve the plane estimation task: a classical solution exploiting the homography constraint (and, thus, almost completely based on image correspondances across distant frames), and an alternative method fully taking advantage of the scene structure estimated incrementally during the camera motion. The two methods are extensively compared in several case studies by discussing the various pros/cons.
Keywords :
cameras; object detection; robot vision; active strategy; homography constraint; monocular camera; optimized camera motion; plane detection; plane estimation; robotic vision; Cameras; Convergence; Eigenvalues and eigenfunctions; Estimation; Three-dimensional displays; Transmission line matrix methods; Vectors;
Conference_Titel :
Robotics and Automation (ICRA), 2014 IEEE International Conference on
Conference_Location :
Hong Kong
DOI :
10.1109/ICRA.2014.6907555