Signature Based Radar Target Classification

Abstract

This study attempts to categorize ten aerial targets, including fighter jets, missiles, military helicopters, and unmanned aerial vehicles (UAVs). A large dataset comprising of simulations of aerial targets at various aspect angles is taken rather than real-time data in order to attain higher accuracy and better classification. This study proposes a highly accurate multi-model radar target classification system that performs a comparative analysis of machine learning and deep learning algorithms such as Random Forests, Support Vector Classifier (SVC), k-Nearest Neighbors (KNN), Convolutional Neural Networks (CNN), Long Short Term Recurrent Neural Networks (LSTM RNN).