Sea-ice velocity fields estimation on Ross Sea AVHRR images

NOAA-AVHRR imagery of the Ross Sea (Antarctica) of December 1990 has been analysed in order to obtain information about the geophysical processes acting during the melting season. A complete methodology is proposed for automatic tracking of sea-ice in daylight AVHRR data. Previous tracking algorithms have shown to be able to retrieve sea-ice motion under limited atmospheric conditions: cloud free data are normally required. In this study weaker hypothesis are introduced in order to remain under quite normal circumstances. Two aspects are specially outlined: the use of partially cloudy monocular images and the estimation of ice pack trajectories along an image sequence. First of all, a classification technique is applied for the discrimination of clouds over the sea and for the detection of snow-ice regions. Then, an optima matching filter is used for the sea-ice motion estimation. The finally derived vector field is homogeneous and shows the ice pack motion along three days image data