SAR Remote Sensing of Canadian Coastal Waters using Total Variation Optimization Segmentation Approaches
MetadataShow full item record
The synthetic aperture radar (SAR) onboard Earth observing satellites has been acknowledged as an integral tool for many applications in monitoring the marine environment. Some of these applications include regional sea-ice monitoring and detection of illegal or accidental oil discharges from ships. Nonetheless, a practicality of the usage of SAR images is greatly hindered by the presence of speckle noises. Such noise must be eliminated or reduced to be utilized in real-world applications to ensure the safety of the marine environment. Thus this thesis presents a novel two-phase total variation optimization segmentation approach to tackle such a challenging task. In the total variation optimization phase, the Rudin-Osher-Fatemi total variation model was modified and implemented iteratively to estimate the piecewise smooth state by minimizing the total variation constraints. In the finite mixture model classification phase, an expectation-maximization method was performed to estimate the final class likelihoods using a Gaussian mixture model. Then a maximum likelihood classification technique was utilized to obtain the final segmented result. For its evaluation, a synthetic image was used to test its effectiveness. Then it was further applied to two distinct real SAR images, X-band COSMO-SkyMed imagery containing verified oil-spills and C-band RADARSAT-2 imagery mainly containing two different sea-ice types to confirm its robustness. Furthermore, other well-established methods were compared with the proposed method to ensure its performance. With the advantage of a short processing time, the visual inspection and quantitative analysis including kappa coefficients and F1 scores of segmentation results confirm the superiority of the proposed method over other existing methods.
Cite this version of the work
Tae-Jung Kwon (2011). SAR Remote Sensing of Canadian Coastal Waters using Total Variation Optimization Segmentation Approaches. UWSpace. http://hdl.handle.net/10012/5893