Underwater Camera Calibration and 3D Reconstruction

  • Author / Creator
    Yau, Timothy HY
  • The recovery of 3D information from 2D images is a well-studied problem in computer vision, with many competing methods that can achieve highly accurate results. However, relatively little attention has been paid to the problem of 3D reconstruction in underwater environments. When cameras are placed underwater they must be protected by a waterproof housing, often featuring a flat glass port through which the scene is viewed. Light rays passing from water into the housing are bent by refraction, a nonlinear process that renders the standard perspective camera model invalid. In spite of this, it is common practice in photogrammetry studies to treat refraction as a radial lens distortion, which can lead to errors in the measured 3D information. Recent research has acknowledged that a physically-correct model of refraction is needed to obtain accurate 3D reconstructions in underwater environments. One important and necessary step is to calibrate the parameters of such a refraction model. In this thesis we develop a novel calibration method that exploits the dispersion of light, which is the angular separation of different wavelengths during refraction, and show that better accuracy is achieved compared to previous work. We then show how to adapt existing reconstruction algorithms to use the physical refraction model, and combine it with our calibration method to obtain a complete process for underwater 3D reconstruction. By analyzing the reconstruction results and comparing against the standard perspective model approximation, we identify the shortcomings of the latter and reveal situations where the refraction model is indispensable. Lastly, we apply the techniques developed in this thesis to real data captured by an underwater observatory and obtain high quality 3D reconstruction results.

  • Subjects / Keywords
  • Graduation date
  • Type of Item
  • Degree
    Master of Science
  • DOI
  • License
    This thesis is made available by the University of Alberta Libraries with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.
  • Language
  • Institution
    University of Alberta
  • Degree level
  • Department
    • Department of Computing Science
  • Supervisor / co-supervisor and their department(s)
    • Yang, Herbert (Computing Science)
    • Gong, Minglun (Computing Science)
  • Examining committee members and their departments
    • Yang, Herbert (Computing Science)
    • Basu, Anup (Computing Science)
    • Gong, Minglun (Computing Science)