Fractal feature based radar signal classification

Automatic target recognition (ATR) using radar, is a much researched area with a demand for better and more accurate classification algorithms. In the present work, the local fractal dimensions of a synthetic aperture radar image have been used as features to classify ground targets. The performance of the fractal feature based ATR algorithm was compared with that of three other established ATR algorithms, viz. the simple yet powerful template matching ATR algorithm, the Gaussian model based Bayesian classifier, and the recently proposed principal component analysis based nearest neighbour algorithm. The fractal feature based classifier was shown to outperform all the other algorithms.