Abstract:
The core focus of the thesis is making a course estimation of a wave buoy using two cameras. The method is based on an initial coarse tracking done individually on each camera using a Kernelized Correlation Filter, followed by triangulation and a multi-view optimization to refine the buoy trajectory. The tracker is paired with an Extreme Learning Machine classifier to detect whether the buoy is occluded, and to help with re-detection and optimization. The challenge it looks to solve is tracking a buoy, especially when faced with partial or full occlusion, and offering assistance in processing sensory data of the wave buoys by giving their exact location. Which has a known problem where the buoy data is measured without the ability to consider the true height of the crest of the tides. We have tested our method on a real dataset, and the results are demonstrated in the thesis, [ which reveal that it is possible to improve the buoy measurements by using these methods]
