Combining morphological mapping and principal curves for ship classification

In this work, we develop a ship classifier, which employs principal curves to extract relevant information from segmented images. This classifier is based on the Euclidean distance of the point whose coordinates represent distinguishing features extracted from an incoming ship image to the principal curve assigned to each class. This methodology is attractive, since it has a low computational cost for the operational phase and easily scales up to an arbitrary number of classes. A mean classification efficiency of 97.3% was achieved, which outperforms previous results based on neural network architecture.