Person Re-identification Based on Human Body Parts Signature

Abstract

In video surveillance, person re-identification consists in recognizing a person or determine whether a given person has already appeared over a large-scale network of cameras. In general, the human appearance obtained in one camera is usually different from the others obtained in other cameras, because of variations in view angle, illumination, body pose, background clutter and occlusion. To address this problem, we propose a person re-identification approach by using fusing multiple human body parts signature method. Firstly we detect human and human body parts. secondly we extract discriminatingly features on each of the parts. Finally, we map the person re-identification issue to a distance learning problem, which aims to find out the similarity between corresponding body parts. The experimental result on several public benchmark datasets (Prid2011,ViPER,ETHZ) shows that the proposed approach works well and leads to a high accuracy of person re-identification.