Automatic detection of sea ice drift from SAR data

Sea ice drift can be automatically determined from a sequence of SAR satellite images. We use a pattern-based multi-resolution cascaded motion tracking algorithm to determine sea ice drift from SAR images. Without field data for validation, it is difficult to assess the accuracy of the derived drift product. The cross-correlation coefficient is often used as a measure for the reliability of the result. It can be shown that it is not a robust criterion for the quality of the drift vectors. We will present an improved quality measure which integrates image statistics and texture as well as a consistency check for the drift estimate and show examples from the Ronne Polynia (Antarctica).