Generating a Rule Base for Sea Ice Classification
Project Award Date: 0000-00-00
The goal of this project is to generate an implementable rule base to assist in the classification of different sea ice types from Synthetic Aperture Radar (SAR) satellite images.
We are interviewing sea ice experts and sea ice observers from the U.S. National Ice Center and the Canadian Ice and Environment Center to generate a rule base for classification. The qualitative rules experts are providing are also matched with quantitative image features and characteristics. For example, "rough-looking ice" is translated by our work into a fuzzy description of various texture parameters that can be computed from the image.
The result is a collection of rules and of image analysis algorithms that generate the predicates and descriptions used in the rules.
Our system is being tested on SAR imagery of the Beaufort Sea and is being evaluated by human experts in the U.S. National Ice Center and the Canadian Ice and Environment Center.
Primary Sponsor(s): Naval Research Laboratory, NASA Headquarters