Person re-identification based on contextual characteristic

An efficient contextual characteristic is proposed for person re-identification. Most current approaches are based on either constructing robust appearance descriptors or learning a distance metric for precise feature matching. However, re-identifying results may be inaccurate and not robust due to appearance features variation caused by various environment changes and individual movement factors. In this reported work consideration is given to the introduction of the contextual characteristic that contains similarities of both k-nearest and ḱ-farthest neighbours between the probe and the gallery, and combines it with Mahalanobis distance for ranking every gallery image more accurately. The experimental result has validated the effectiveness of the proposed method on a challenging publicly available dataset.