Automatic estimation of camera parameters from a solid calibration box

  • PDF / 389,687 Bytes
  • 9 Pages / 595.276 x 790.866 pts Page_size
  • 103 Downloads / 296 Views

DOWNLOAD

REPORT


TECHNICAL PAPER

Automatic estimation of camera parameters from a solid calibration box P. R. G. Kurka • J. V. Delgado • C. R. Mingoto O. E. R. Rojas



Received: 12 August 2011 / Accepted: 14 March 2012 / Published online: 11 April 2013 Ó The Brazilian Society of Mechanical Sciences and Engineering 2013

Abstract The paper presents a simple and automatic procedure to estimate the intrinsic and extrinsic parameters of a camera, based on the image of a standard calibration box. Image processing techniques are used to determine the relevant image points to be used in the calibration procedure. A non-recursive solution scheme is proposed to estimate the intrinsic and extrinsic camera parameters. Simulated and real applications are presented to illustrate the use and performance of the proposed technique. Keywords geometry

Camera calibration  Camera model  Vision

1 Introduction Calibration is a necessary operation in photogrammetric applications. It consists in finding intrinsic and extrinsic parameters of a pinhole camera model. The extrinsic parameters are used in the conversion of 2-D projected images into real 3-D world objects. Extrinsic parameters are

Technical Editor: Glauco Carin. P. R. G. Kurka (&)  J. V. Delgado  C. R. Mingoto  O. E. R. Rojas Faculdade de Engenharia Mecaˆnica, Universidade Estadual de Campinas, 055, Campinas, Sa˜o Paulo, Brazil e-mail: [email protected] J. V. Delgado e-mail: [email protected] C. R. Mingoto e-mail: [email protected] O. E. R. Rojas e-mail: [email protected]

related to the movement of a camera, from the image of a regular object of known dimensions, with respect to a reference frame. Such parameters can also be used to determine the dimension and position of objects in space, with respect to an arbitrary referential, from their simultaneous images in a set of two or more cameras. Dimensional information of objects in space from their camera projections can only be obtained after the intrinsic parameters of the camera have been determined. Intrinsic parameters consist of the camera’s focal length, expressed in horizontal and vertical pixel dimensions, skew parameter of the camera’s sensing element and the pixel coordinates of the intersection of the camera’s optical axis with the image plane. Some popular camera calibration techniques, such as [1, 2], require several images of planar calibration grids, and user actions to indicate special points in the displayed images. The constitutive algebraic equations of such applications require the use of iterative optimization solutions, such as the maximum likelihood scheme, to find homographies between the calibration grid points and its images. The present work proposes a calibration procedure based on the processing of edge points of the single image of an ordinary rectangular box of known dimensions, found in the context of any image processing laboratory. The method uses the concepts of perspective projection and vanishing lines, and is a simple and robust tool to estimate the intrinsic and extrinsi

Data Loading...