Radar Target Recognition and Classification Using Supervised Machine Learning Approaches

Abstract

Target classification from the returned echo signals is one of the challenging problems in modern RADAR systems. The key feature that is used for target classification is the Radar Cross-Section (RCS). The recent advancements in the field of machine learning techniques gave interesting results for the RADAR target recognition. Dedicated machine learning models are realized to recognize simple and complex targets. The models corresponding to both simple and complex target recognition are developed with a capability to identify four common geometrical structures, namely circular cylinder, frustum (truncated cone), circular disk, and sphere. The proposed method extracts features of simple targets by using Maximal Overlap Discrete Wavelet Packet Transform (MODWPT). The unknown targets are classified with the feature extraction set obtained using different supervised classifiers, namely k-Nearest Neighbor (k-NN), Support Vector Machine (SVM), Artificial Neural Network (ANN), and their performance is compared. The k-NN classifier gives better performance of classification accuracy when compared to existing methods.KeywordsRadar cross-sectionMaximal overlap discrete wavelet packet transformk-nearest neighborSupport vector machineArtificial neural network