Feature extraction methods in person re-identification system: a technical review

Abstract

Intelligent surveillance is an emerging research area in the field of computer vision. Person re-identification is one among the tools involved in intelligent surveillance. Person re-identification is used to recognize and identify a person of interest captured by different surveillance cameras at different times and at different locations, when an input image is given. Automation of person re-identification is difficult in real time due to changes in pose, background, illumination and occlusion. Recent researchers have focused on developing discriminant, robust features, learning distance metric models or fusion of both for matching between the images of person. Our main objective is to provide the future researchers the importance of various state-of-the-art feature extraction techniques and deep learning approaches used in person re-identification, till date. Different algorithms with their strengths and accuracy percentage were summarized in a comparison table. Finally, unsolved problems in person re-id were listed that can be used as guidelines for future research.