Marine Target Characteristics in Satellite SAR Imagery
In this thesis, marine targets in Synthetic Aperture Radar (SAR) imagery are studied. The appearance of different marine targets in different polarization channels, their scattering behaviour, and their contrast measures and geometric properties are investigated. RADARSAT-2 data containing ships, oil rigs and icebergs in ocean and/or sea ice background are evaluated. To be able to detect a target in a SAR image, a significant contrast between target and background is needed. In this thesis, peak-to-background ratios (PBR) and target-to-background ratios (TBR) are used to evaluate the contrasts for different target types and background classes. The relations between contrast measures and incidence angle, target size and weather conditions are addressed. In order to track, identify or classify targets, reasonable features for discrimination must be chosen. In this project, Hu's 1st geometric moment is used as a geometric feature to investigate the elongatedness of the targets. It is found that cross-polarization (HV and VH) gives better contrast than co-polarization (HH and VV), especially at low incidence angles. Where quad-polarimetric data is available, enhanced contrast can be obtained through polarimetric decompositions. For man-made targets, the contrasts are seen to be dominated by the volume scattering and double bounce components, i.e. HV+VH and HH-VV. For man-made targets in ocean, contrasts for co-polarization channels are seen to increase with incidence angle, while cross-polarization contrasts decrease. The contrast measures are also seen to increase with target size, and to decrease with increased wind in the case of ocean background. Comparison of contrasts for different target types shows that PBR is larger for man-made targets in sea ice than in ocean areas, while the opposite is seen for TBR. Comparison of man-made targets versus icebergs shows that the latter has highest mean PBR while man-made targets have highest mean TBR. Ships in sea ice have higher contrasts than icebergs in both PBR and TBR. More data would provide more reliable results and a more accurate comparison. It is seen that Hu's 1st moment may be a useful feature for distinguishing between small and large ships, or between larger ships and other target types.
ForlagUniversitetet i Tromsø
University of Tromsø
MetadataVis full innførsel
Følgende lisensfil er knyttet til denne innførselen: