Robust ISAR image classification using Abridged Shape Matrices

In this paper a new classification method is proposed to classify the ISAR images automatically based on Abridged Shape Matrices (ASMs). Initially, a considerate noise removal procedure is used to abate the noise present in the ISAR images and the target is segmented out from the background clutter. ASMs are obtained from the segmented targets by careful interpolation of the shape details of the target by the polar quantization. ASMs are concise and non redundant as the over sampling is avoided in interpolating the inner shape details of the target. The avoidance of superfluous interpolations in polar quantization reduces the generation time of the ASMs, which in turn curtails the over all execution time of the classifier. Experiments were conducted on synthesized ISAR aircraft image dataset and results are presented. The experimental results show that the proposed method is sturdy and more meticulous than existing methods.