Research Progress and Trend of Person Re-identification
Download as PDF
DOI: 10.23977/cii2019.46
Corresponding Author
Guankun Wang
ABSTRACT
Person re-identification refers to the matching technology of the same pedestrian image under different non-overlapping cameras, which has important application value in strengthening social management, preventing the occurrence of criminal acts and realizing event reconstruction. Person re-identification mainly relies on human visual representation and artificial design features and is greatly influenced by illumination, image resolution, pedestrian posture and shooting angle. Therefore, person re-identification faces great challenges. In this paper, the existing pedestrian representation feature learning technology and measurement technology are reviewed and analyzed, and the existing problems and possible solutions are pointed out. Person re-identification has great significance for researchers in this field to grasp the status quo and put forward new research ideas.
KEYWORDS
Person re-identification, image features, deep learning, semantic information