Machine Learning Approach for Efficient Radar Target Classification

This study investigates radar-based detection and classification of cylindrical and cone-shaped targets using various fractures of the moving targets. Cylindrical targets demonstrate a consistent radar cross-section (RCS) profile with a peak of 25 m<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> and a maximum detection range of 31.65 km, while cone-shaped targets exhibit angle-dependent reflections with a peak RCS of 15 m<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> and a reduced detection range of 26.50 km. Doppler shift analysis provides target speed ranges of 8.82–44.11 m/s for cylinders and 5.29–26.47 m/s for cones under varying motion conditions. A system using machine learning is created and tested with Support Vector Machine (SVM), Long Short-Term Memory (LSTM), and SqueezeNet models to check how well it can identify different targets. The classification accuracies achieved are 97% (SVM), 98% (SqueezeNet), and 99.3% (LSTM), clearly outperforming conventional target classification techniques. These results confirm the effectiveness of the proposed approach in distinguishing between geometrically distinct targets under ideal conditions.