HRRP SCATTERING CENTER ESTIMATION BASED ON MODELLED NEURAL NETWORK

Scattering center estimation of HRRP is crucial for radar automatic target recognition (RATR). Traditional estimation algorithms either necessitate prior information of the signal or suffer from mismatch issues or possess excessive computational costs. In this paper, we propose an HRRP-modelled neural network (HNN), which addresses the mismatch problem and exhibits better accuracy. HNN combines the HRRP signal model with its layered structure, including input, output and activation function. It is initialized via orthogonal matching pursuit and optimized using back propagation algorithm with dual learning rate. The estimated position and amplitude can be obtained by the weights between layers. Through simulations and experiments, we demonstrate the superiority of HNN over traditional methods.