Person Re-Identification Using MmWave Radar

Abstract

In this study, person re-identification was performed using a millimeter-wave radar. Camera and radar data were processed separately using deep learning methods, and the results were compared. Although camera-based systems produce highly accurate results, they also pose significant privacy concerns. The low computational requirements of millimeter-wave radar and its robustness to environmental conditions make it advantageous compared to traditional camera systems. In this study, raw radar data were transformed into range-Doppler maps using a fast Fourier transform, and these data were analyzed using ResNet-101 and a long short-term memory network. Additionally, the person recognition performance of radar data was evaluated in comparison with camera-based systems. The results demonstrate that radar-based systems can serve as a secure, privacy-friendly, low-cost, and efficient alternative for person re-identification. © 2025 Elsevier B.V., All rights reserved.