A Review of Research on Cross Domain Pedestrian Re-identification Based on Deep Learning

Pedestrian re-identification (Re-ID) is crucial in urban safety and traffic surveillance. However, research on cross-domain techniques is particularly necessary due to challenges such as ambient lighting, camera viewpoints, and background interference. This article first provides an overview of the foundation of pedestrian re identification. Subsequently, the application of two state-of-the-art technologies, Domain Adaptation (DA) and Generative Adversarial Networks (GANs), in cross domain pedestrian re identification and their recent research progress were discussed in depth. Next, the principal metrics for assessing the effectiveness of pedestrian Re-ID are thoroughly explained, and the experimental performance of these techniques on major classical datasets is compared. Finally, the challenges faced by the current techniques and the possible directions for future research are summarised and discussed.