Calibration of an underwater stereoscopic vision system

Although the problem of terrestrial camera calibration has been studied extensively, this knowledge does not necessarily transfer to the underwater environment directly. Because of the significant differences between the optical properties of the two transfer media, it is necessary to address the task of underwater camera calibration as a unique problem. Correspondingly, this paper studies the differences between terrestrial and underwater camera calibration and shows that the calibrated camera models are significantly different in these two environments. Thus, the necessity for in-situ calibration for an underwater environment is quantitatively ascertained. In addition, an underwater stereoscopic vision system is calibrated employing two calibration algorithms; namely, the Rahman-Krouglicof and the Heikkila algorithm. Since the mathematical formulations of the calibration problem in both environments are identical, a general calibration algorithm can be adopted for an underwater application provided that it is robust enough to overcome the suboptimal imaging conditions of an underwater environment. In order to identify a suitable calibration algorithm for aquatic environments, this paper assesses the stereoscopic performance of the two calibration algorithms in terms of reconstruction error. The experimental data confirms that the Rahman-Krouglicof algorithm is well-equipped to address the peculiarities of underwater 3D reconstruction.