Encoding the regional features for person re-identification using locality-constrained linear coding
Junnan Li, Zhen Yang, Huilin Xiong · 2015
This paper presents a coding method for person re-identification, based on Locality-constrained Linear Coding (LLC), utilizing the locality constraints to project each descriptor into its local-coordinate system, in which a spatially weighted color feature combining with the HoG descriptor is introduced. The proposed encoding feature for LLC retains the basic form of Iterative re-weighting Sparse Ranking (ISR), and the final classification is directly made according to a normalized reconstruction error. Compared with other approaches, the proposed method avoids extracting sophisticated features and does not require the learning of a classifier. Experimental results show that the proposed approach improves rank-1 recognition accuracy on the CAVIAR4REID dataset from 29% to 31% with much lower computational costs.