A robust and efficient SAR ATR algorithm using a hybrid model of fractional fourier transform and pulse coupled neural network

Abstract

A hybrid framework consisting of Fractional Fourier Transform (FrFT) and Pulse coupled Neural network (PCNN) is proposed in this paper for highly accurate and orientation, position & scale invariant synthetic aperture radar (SAR) automatic target recognition (ATR). FrFT is used to gather scattering information and insights that are attainable using timefrequency and time-scale techniques, whereas PCNN is used to achieve invariant target recognition. Public release of the MSTAR dataset is used to validate the proposed system. We compared our proposed system performance with existing approaches and established the better performance of this system. We have shown that, even with reduced training sets, the proposed system shows consistent performance whereas the performance of conventional systems degrades.