SAR Target Classification Enhancement via Synthetic Defocused Image Augmentation Using Kinematic Motion Modeling

Deep neural networks (DNNs) have demonstrated remarkable potential in automatic target classification using synthetic aperture radar (SAR) imagery. However, their performance heavily relies on the availability of sufficient training data, which is often limited in practice particularly for moving targets. To address this challenge, data augmentation becomes indispensable for enhancing classification robustness and generalization. This paper proposes a novel augmentation technique that generates defocused SAR images by inserting kinematic motion-induced phase distortions into focused SAR data. Target motion is mathematically modeled using a polynomial-based trajectory representation, from which a phase modulation filter is derived to simulate the defocusing effect. Experimental results on the real dataset demonstrate that incorporating the augmented defocused images significantly improves target classification performance of DNN-based classifiers compared to training with original data alone.