Abstract

Most existing person re-identification methods focus on find- ing similarities between persons between pairs of cameras (camera pair- wise re-identification) without explicitly maintaining consistency of the results across the network. This may lead to infeasible associations when results from different camera pairs are combined. In this paper, we pro- pose a network consistent re-identification (NCR) framework, which is formulated as an optimization problem that not only maintains consis- tency in re-identification results across the network, but also improves the camera pairwise re-identification performance between all the indi- vidual camera pairs. This can be solved as a binary integer programing problem, leading to a globally optimal solution. We also extend the pro- posed approach to the more general case where all persons may not be present in every camera. Using two benchmark datasets, we validate our approach and compare against state-of-the-art methods.