Automatic target recognition of aircrafts using translation invariant features and neural networks

Automatic target recognition (ATR) of aircrafts using translation invariant features derived from high range resolution (HRR) profiles and multilayered neural network is presented in this paper. The HRR profile sequences are translation variant in the range resolution cell because of the non-cooperative target maneuvering. The differential power spectrum (DPS) is introduced to extract the translation invariant features. Several learning algorithms of feed-forward neural network are implemented to determine an optimal choice in the recognition phase. The range profiles are obtained using the two-dimensional backscatters distribution data of four different scaled aircraft models. Simulations are presented to evaluate the classification performance with the DPS based features and neural networks. The results show that this method is effective for the application of radar target recognition.