Radar HRRP Signal Re-ID via Deep Learning

High-resolution range profile (HRRP) is often used in radar automatic target recognition (RATR) tasks. The performance of HRRP classification has received promising results with convulotional neural networks (CNN). However, in real-world application, it is difficult to obtain sufficient HRRP data with all aspect-angles for training the networks, especially for non-cooperative targets. Such insufficient HRRP data dramatically hampers classification performance. Thus, the proposed work introduces re-identification (ReID) into the HRRP recognition. A suitable model based on CNN for HRRP signal is proposed in the proposed work. Experiments on public datasets demonstrated that the proposed methods outperform the CNN-based classification model in target recognition.