A Data Driven Approach for Target Classification Based on Histogram Representation of Radar Cross Section

Abstract

A new approach for classifying targets based on their radar cross section (RCS) is discussed. The RCS presents unique statistical features depending on the target’s shape, while an incident angle with small random fluctuation is considered. Data sets are generated utilizing Physical Optics simulation of the RCS, and the classification of targets with different shapes is performed by Artificial Neural Network (ANN). The algorithm’s performance is evaluated, especially regarding the robustness against noise on the RCS data. Numerical examples motivated by mm-wave radar applications in driving assistance systems are presented. The results show that the classification algorithm performs promising results and ensures the robustness of the features extracted from histogram definitions of RCS.