People re-identification using 3D descriptor with skeleton information

People re-identification is a fundamental task for automated video-surveillance applications and has attracted attention of many researchers in past few years. Most of the studies in this field are based on 2D images and color information. In these methods it is assumed that the individuals do not change their clothes, thus these methods cannot be used for long term re-identification. To overcome this problem, we proposed a novel approach for people re-identification based on 3D information. In this paper we used a combination of 3D descriptors of body shape and skeleton data which is independent of the clothes color and illumination changes and it can be used for long term re-identification. We evaluated our work on the state-of-the-art RGB-D dataset BIWI. The results of this evaluation show that the proposed method achieved high performance in comparison to some recent methods.