Person re-identification using kNN classifier-based fusion approach

Re-identification is the process of identifying the same person from images or videos taken from different cameras. Although many methods have been proposed for re-identification, it is still challenging because of unsolved issues like variation in occlusions, viewpoint, pose and illumination changes. The objective of this paper is, to propose a fusion-based re-identification method to improve the identification accuracy. To meet the objective, texture and colour features are considered. In addition the proposed method employs Mahalanobis metric-based kNN classifier for classification. The performance of proposed method is compared with the existing feature-based re-identification methods. CAVIAR, VIPeR, 3DPes, PRID datasets is used for experiment analysis. Results show that the proposed method outperforms the existing methods. Further it is observed that Mahalanobis metric-based kNN classifier improves the recognition accuracy in re-identification process.