Recent Advances in Person Re-identification Based on Deep Learning

Person re-identification can identify and match persons with the same identity as the target person in cross-camera scenarios, providing strong technical support for traffic management, public safety, smart city construction, etc. Aiming at the practical application needs of person re-identification, this paper first systematically introduces and summarizes person re-identification, including the research status, datasets, and evaluation indicators of person re-identification. Afterwards, this paper summarizes and analyzes the cutting-edge research in three areas: unsupervised person re-identification, cloth-changing person re-identification, and occluded person re-identification, while their current development status is summarized, and the existing methods in each direction are sorted out and compared. Finally, the development trend of person re-identification is analyzed and prospected. This paper reviews person re-identification methods, hoping to provide reference and help for researchers to further carry out relevant research in the field of person re-identification and promote the development of person re-identification technology.