Abstract
An algorithm is presented for the stabilization and rectification of digital video from floating platforms. The method relies on a horizon-tracking technique that was tested under a variety of lighting and sea-state conditions for 48 h of video data over 12 days during a research cruise in the North Pacific Ocean. In this dataset, the horizon was correctly labeled in 92% of the frames in which it was present. The idealized camera model assumes pure pitch-and-roll motion, a flat sea surface, and an unobstructed horizon line. Pitch and roll are defined along the camera look direction rather than in traditional ship coordinates, such that the method can be used for any heading relative to the ship. The uncertainty in pitch and roll is estimated from the uncertainties of the horizon-finding method. These errors are found to be of the order 0.6° in roll and 0.3° in pitch. Errors in rectification are shown to be dominated by the uncertainty in camera height, which may change with the heave motion of a floating platform. The propagation of these errors is demonstrated for the breaking-wave distribution Λ(c). A toolbox for implementation of this method in MATLAB is shared via the MATLAB File Exchange.
Supplemental information related to this paper is available at the Journals Online website: http://dx.doi.org/10.1175/JTECH-D-14-00047.s1.
