Person re‐identification by multi‐statistics on pyramid of covariance matrices

Abstract

A novel and efficient covariance-based method for person re-identification is proposed. The approach exploits three colourspaces and intensity gradients as covariance features and extracts multiple statistical feature vectors from the pyramid of region covariance matrices. The distance measure of the covariance pyramid is designed to be the weighted combination of four vectorised statistical features by cascading on the covariance pyramid. The method is compared with the state-of-the-art methods using a benchmark dataset and is demonstrated to outperform other state-of-the-art methods.