Image-Analysis Techniques for Determination of Morphology and Kinematics in Arctic Sea Ice

Abstract

The synthetic aperture radar (SAR) data have been used to study sea ice with respect to its motion and formation/deformation. With the prospect of the Alaska SAR Facility development in the near future, there is a great need for robust and efficient sea-ice analysis techniques. This paper presents a sea-ice motion analysis technique that can be used for (1) local motion analysis of a selected ice patch and (2) a global ice motion over the entire image area. Though there are several sea-ice motion tracking techniques, they do not provide the required operational speed or robustness. In order to meet the operational speed requirement (over fifty images per day), we have developed a sea-ice motion analysis technique which requires very little human interaction and much simplified computation. The proposed technique uses a subset of easily distinguishable features to predict global motion characteristics and apply template matching over a predicted search area. We applied the developed technique to two pairs of SEASAT SAR images, one pair with a minor motion of “ice pack” and another with a larger and discontinuous motion of “fast ice”. The two major achievements of the new approach are: first, development of a set of computer-aided tools for feature selection and registration and, secondly, implementation of an optimal search strategy for automatic template matching via a motion prediction model.