INVESTIGATING THE POTENTIAL OF DIFFERENT POLARIMETRIC FEATURES BASED ON DUAL POLARIMETRIC TERRASAR-X DATA FOR AUTOMATED SEA ICE CLASSIFICATION

摘要

In this work, we examine the performance of an automatedrnsea ice classification algorithm based on dual polarimetricrnTerraSAR-X data. Polarimetric features are extractedrnfrom HHVV dualpol stripmap images. In a secondrnstep, the feature vectors are fed into an artificial neuralrnnetwork to classify each pixel into an ice type. Thernfirst part of our analysis addresses the predictive valuernof different subsets of features for our classification processrn(by means of measuring mutual information). Differentrnneural network configurations are then exploredrnfor optimal classification performance. The results on arnTerraSAR-X dataset indicate a high reliability of a trainedrndual polarimetric classifier. Performance speed and accuracyrnpromise applicability for near real time operationalrnuse.