17 / 17
Page 62
Journal of Oceanography and Marine Research | ISSN: 2572-3103 | Volume 6
&
Marine Science, Coastal Dynamics and Management
6
th
International Conference on
Oceanography, Ocean Technology and Marine Biology
6
th
International Conference on
September 21-22, 2018 | Dallas, USA
Camera calibration: What is important and what is important for underwater imagery
Yuri Rzhanov
University of New Hampshire, USA
C
amera calibration is an essential part of image processing pipeline resulting with a 3D scene reconstruction. Traditionally, this
term encompasses determination of properties of individual cameras (intrinsic parameters) and calculation of mutual poses of
cameras (extrinsic parameters). Processing of underwater imagery adds another layer of complexity related to waterproof housings
that enclose cameras, with transparent windows refraction index of which is different from that of air and water. Thus, light rays
reflecting from an imaged object bend two times before arriving at a sensor–at water/window and window/air interfaces. For
commonly used flat or dome-shaped windows, accounting for refraction adds three more calibration parameters. Until recently,
many researchers were attempting to describe refractive distortion in terms of radial lens distortion, but this approach was proven to
be not viable, as in presence of refraction camera becomes varifocal. Any quantitative result obtained from image processing requires
knowledge of uncertainty related to calibration accuracy. We assume that resources available to researchers–measured in monetary
and time units–are limited, and thus the guidance how to achieve maximal accuracy with minimal cost is important. This paper
presents a comprehensive investigation of factors affecting the accurate determination of all calibration parameters. These factors
include pixelation noise, shape, and pattern drawn on a calibration object, the necessary number of acquired images, the structure
of a multi-camera rig, etc. For underwater cameras, with refractive interfaces, we propose a novel technique for determination
of refraction-related parameters. This technique is fast and reliable, as it requires the only optimization in one dimension for flat
interfaces. Experimental results demonstrate a good correlation between calculated parameters and the parameters that were possible
to measure manually. Accurate knowledge of refractive calibration parameters allow for incorporation of refraction effects in a 3D
scene reconstruction workflow and dramatically improves reconstruction results.
yuri.rzhanov@unh.eduJ Oceanogr Mar Res 2018, Volume 6
DOI: 10.4172/2572-3103-C1-006