Sea state from monoscopic ocean video in real environments
Loizou, A
Date: 8 November 2021
Publisher
University of Exeter
Degree Title
PhD in Computer Science
Abstract
Video of the ocean surface is used as a means for estimating useful information about the scene. A methodology is first introduced for approximating the pixel to metre scale from high-scale videos of the ocean, such as from an aeroplane. Radar images are used for testing. The temporal and spatial domains are associated through the phase ...
Video of the ocean surface is used as a means for estimating useful information about the scene. A methodology is first introduced for approximating the pixel to metre scale from high-scale videos of the ocean, such as from an aeroplane. Radar images are used for testing. The temporal and spatial domains are associated through the phase modulation of waves, and a process is introduced that selects the waves with the highest energy to be used for estimating the pixel scale. The spatial information is then used with the calculated pixel scale for approximating the sea state.
Due to the difficulty of obtaining high-scale videos, a methodology is then introduced that uses the temporal variation from video, and specifically time series of pixel intensities. It aims to isolate and utilise the temporal variation of the wave field from all other video elements, such as environmental brightness fluctuations. The methodology utilises the Kalman filter and the least squares approximate solution for providing an uncalibrated video amplitude spectrum. A method is proposed for scaling this spectrum to metres with the use of an empirical model of the ocean. The significant wave height is estimated from the calibrated video amplitude spectrum. Videos of the ocean in real environments from a shipborne camera and a tower are used for testing. In both sets of data, in situ buoy information is used solely for validation.
The next technique aims to approximate the sea state from the same kind of data, namely videos of the ocean in real environments, without calibrating a video amplitude spectrum. The proposed methodology tracks the principal component of the movement of water in the video, which is speculated to be associated with the dominant frequency of the ocean. To accomplish this, the singular spectrum analysis algorithm and the extended Kalman filter are used. Then, the shape of an empirical spectrum is utilised in order to translate the dominant frequency output into a significant wave height estimation.
The problem of not using ocean theory associated with a particular empirical energy spectrum for calibration is examined in the next methodology. A secondary oscillatory component from the singular spectrum analysis algorithm is identified with the incorporation of the extended Kalman filter. Ocean theory involving the equilibrium range of oceans is used for calibration.
The shipborne videos are used for testing the behaviour of the techniques for approximately the same sea state of 3.1m to 3.4m of significant wave height. The tower videos are used for testing the techniques for a variety of sea states ranging between 0.5m and 3.6m of significant wave height. From all methodologies, the maximum observed values of root mean square error 0.37m and of mean absolute percentage error 18% suggest that the work is promising at estimating these states.
Doctoral Theses
Doctoral College
Item views 0
Full item downloads 0