Object Pose: The Link between Weak Perspective,Paraperspective, and Full Perspective

  • Authors:
  • Radu Horaud;Fadi Dornaika;Bart Lamiroy

  • Affiliations:
  • GRAVIR-IMAG &/ INRIA Rh\'hat{o}ne-Alpes, 655 ave de l‘/Europe, 38330 Montbonnot, France/ E-mail: First.Last@imag.fr;GRAVIR-IMAG &/ INRIA Rh\'hat{o}ne-Alpes, 655 ave de l‘/Europe, 38330 Montbonnot, France/ E-mail: First.Last@imag.fr;GRAVIR-IMAG &/ INRIA Rh\'hat{o}ne-Alpes, 655 ave de l‘/Europe, 38330 Montbonnot, France/ E-mail: First.Last@imag.fr

  • Venue:
  • International Journal of Computer Vision
  • Year:
  • 1997

Quantified Score

Hi-index 0.00

Visualization

Abstract

Recently, DeMenthon and Davis (1992, 1995) proposed a method for determining the pose of a 3-D object withrespect to a camera from 3-D to 2-D point correspondences. Themethod consists of iteratively improving the pose computed with aweak perspective camera model to converge, at the limit, to a poseestimation computed with a perspective camera model. In this paperwe give an algebraic derivation of DeMenthon and Davis‘ method and weshow that it belongs to a larger class of methods where theperspective camera model is approximated either at zero order (weakperspective) or first order (paraperspective). We describe in detailan iterative paraperspective pose computation method for both noncoplanar and coplanar object points. We analyse the convergence ofthese methods and we conclude that the iterative paraperspectivemethod (proposed in this paper) has better convergence propertiesthan the iterative weak perspective method. We introduce a simple wayof taking into account the orthogonality constraint associated withthe rotation matrix. We analyse the sensitivity to camera calibrationerrors and we define the optimal experimental setup with respect toimprecise camera calibration. We compare the results obtained withthis method and with a non-linear optimization method.