High-Point Descriptor Technique for the Identification of Ship Silhouettes at Various Aspect Angles.

A High-point descriptor technique is developed for the identification of binary grey level silhouette images of ships. These descriptors are theoretically invariant to translation, dilation and rotation and this is verified. A High-point descriptor comprising only two characteristics is shown to be sufficient to adequately characterize a ship silhouette. Euclidean distance is used to compare shapes and an identification threshold is defined, above which a high probability of identification occurs. The technique compare favourably with fourier descriptor techniques.