A Performance Analysis of Neural Network Models in HRRP Target Recognition

In recent years, deep learning method represented by neural network has made great progress in the field of image recognition, but it has not been widely used in the field of high range resolution profile (HRRP) target recognition. On this basis, representative autoencoder (AE), convolutional neural network (CNN) and their derived network structures are applied to HRRP target recognition. Self-built data set is used to identify four types of ship targets. According to the identification accuracy, loss curve, model size and operation time, the comprehensive performance of different models was compared and analyzed, which verified the superiority of neural network model in HRRP target identification, and provided reference for the follow-up work in this field.