Open Access
REVIEW
A Review of Person Re-Identification
Tong Jiang*
Nanjing University of Information Science and Technology, Nanjing, 210000, China
* Corresponding Author: Tong Jiang. Email:
Journal of New Media 2020, 2(2), 45-60. https://doi.org/10.32604/jnm.2020.09823
Received 05 August 2020; Accepted 19 August 2020; Issue published 21 August 2020
Abstract
Recently, person Re-identification (person Re-id) has attracted more and
more attention, which has become a research focus of computer vision community.
Person Re-id is used to ascertain whether the target pedestrians captured by cameras
in different positions at different moments are the same person or not. However, due
to the influence of various complex factors, person Re-id still has a lot of holes to be
filled. In this paper, we first review the research process of person Re-id, and then,
two kinds of mainstream methods for person Re-id are introduced respectively,
according to the different types of training data they used. After that, we introduce
some specific methods for different kinds of person Re-id, including handcrafted
feature descriptor and metrics learning based methods as well as neural network
based methods. Then, some commonly used datasets and their performance
evaluation criteria are introduced. Finally, we compare these methods in order to
display their advantages and disadvantages. Last but not list, depending on the
current research status and development tendency, we make a prospect for person
Re-id research.
Keywords
Cite This Article
T. Jiang, "A review of person re-identification,"
Journal of New Media, vol. 2, no.2, pp. 45–60, 2020. https://doi.org/10.32604/jnm.2020.09823