Comparison of Feature Based Segmentation of Sar Satellite Sea Ice Images with Manually Drawn Ice Charts

This paper explores automatic segmentation of synthetic aperture radar (SAR) satellite images of sea ice. The investigation is based on a comparison of an automatically segmented SAR image and manual classifications by ice service analysts. The automatic algorithm uses statistical properties of the backscattered radar signals to segment the SAR image into a given number of classes. The numberof classes is determined from available ground truth. Sea ice experts, aided by various in-situ data, were able to label most of the segments from the automatic algorithm. We utilized the physical information in the polarimetric features used in the classification algorithm, in order to further explore the class labeling.