With the advent of airborne and spaceborne synthetic aperture radar (SAR) systems, sea ice classification from SAR images has become an important research subject. Since gray tone alone has proven to be of limited capability in differentiating ice types, texture has naturally become an attractive avenue to explore. Accordingly, performance of texture quantification parameters as related to their ability to discriminate ice types has to be examined. SAR image appearance depends on radar parameters involved in the image construction procedures from the doppler history record. Therefore the feasibility of using universal texture/ice type relationships that hold for all combinations of radar parameters also has to be investigated. To that end, imagery data from three different SAR systems were used in this study. Five conventional texture parameters, derived from the gray level co-occurrence matrix (GLCM), were examined. Two of them were modified to ensure their invariant character under linear gray tone transformations. Results indicated that all parameters were highly correlated. The parameters did not, in general, vary with the computational variables used in generating co-occurrence matrices. Ice types can be identified uniquely by the mean value of any texture parameter. The relatively high variability of texture parameters, however, confuses ice discrimination, particularly of smoother ice types. Ice classification was conducted using a per-pixel maximum likelihood supervised scheme. When texture was combined with gray tone, the overall average classification accuracy was improved. Texture was successful in improving the classification accuracy of multiyear ice but was less promising in discriminating first-season ice types. The best two GLCM texture parameters, according to the computed overall average classification accuracies, were the inverse difference moment and the entropy. A brief description of GLCM texture parameters as related to ice's physical characteristics is presented.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.