A comprehensive survey on person re-identification approaches: various aspects

Abstract

Person re-identification (Re-ID) is an application of video surveillance and has become popular among Computer Vision and Image processing research communities since last decade due to having its strong safety and security potential. It is the process of identifying a person of interest in distributed non-overlapping camera views. Person re-identification has broad application in maintaining the security by re-identifying the malicious persons in networking cameras. Now a days terrorist and criminal activities are increasing day by day and it is utmost important to re-identify a person of interest at public places like � shopping malls, railway stations, airports, huge public events etc. A lot of challenges are involved in the re-identification process like variation in lighting condition, different poses and viewpoints, blurring effect, image resolution, background changes etc. Basically 2 types of datasets (image based, video based) are designed for re-identification purpose based on application and approaches. This paper includes the study of many popular datasets like ViPER, iLIDS, Market1501, DukeMTMC4ReID, CUHK01, CHUK02, CHUK03, PRID2011 etc. including the various parameters (no of persons, no of images, no of cameras, size of frames etc.) and challenges involved in that. In this paper various aspects of person re-identification approaches are discussed including temporal, spatial, feature, distance metric, machine learning, automation etc. to get the comprehensive and exhaustive idea of person re-identification methods.