Deep Learning of Raw Radar Echoes for Target Recognition

Abstract

Synthetic aperture radar (SAR) based classification approaches are commonly used methods for automatic target recognition. However, SAR imaging requires complex two-dimensional matched filtering and interpolation algorithms. In this paper, we propose deep learning technology for automatic target recognition based on raw radar echoes instead of SAR images. A modern convolutional neural network (CNN) model is trained directly by radar-echo training data set, and is evaluated on the testing data set. The experimental results show that the proposed method could achieve high accuracy and efficiency for the target recognition.