Tracking the motion of recognizable sea-ice objects from coastal radar image sequences

摘要

The Finnish Meteorological Institute has installed image-capturing devices on some Baltic Sea coastal radars for operational sea-ice monitoring and ice product validation. These devices produce radar images, which are saved operationally at about every 2min. These data can efficiently be utilized in automated tracking of ice motion over sequences of radar images. Reliable estimates of point-wise ice drift can be used as virtual drifter buoys to validate fine-scale ice models. For this purpose we have developed an algorithm, which first locates objects that can reliably be recognized from one radar image to another, and then tracks the motion of these objects until they are lost by the algorithm. The recognizable objects in the first image of an image sequence are located by requiring an object to include a textural content, i.e. the object does not solely consist of a uniform area, and detected edge corner points. The corner points are required to exclude straight linear edges. After locating a suitable number of traceable objects, the tracking is performed between each pair of successive images using a two-resolution phase-correlation algorithm. We have tested the tracking algorithm using image sequences of two coastal radars collected during the 2010/11 and 2011/12 winters.