An artificial neural network approach to radar target identification in polarization domain

The state-of-the-art of radar target identification (RTI) in the polarization domain is reviewed first, then the possibility of using an artificial neural network to solve the problem of directly extracting the polarization-invariant features is discussed. A modified backpropagation algorithm for a multilayer feedforward neural network is proposed for the case of large training samples. A new method for RTI in the polarization domain using single-frequency multipolarization is proposed. By using a dumbbell target as a simulation model, an experiment is performed which shows that the proposed method in this paper is practicable and effective, and it has a high correct classification rate.<<ETX>>