Summary: | Human re-identification is the core task in most surveillance systems and it is aimed at matching human pairs from different non-overlapping cameras. There are several challenging issues that need to be overcome to achieve reidentification, such as overcoming the variations in viewpoint, pose, image resolution, illumination and occlusion. In this study, we review existing works in human re-identification task. Advantages and limitations of recent works are discussed. At the end, this paper suggests some future research directions for human re-identification.
|