Abstract: In order to overcome the inadequate consideration of the existing dictionary learning taken into the con- nection of pedestrian features of different camera views,a new pedestrian re-identification method is proposed on the basis of dictionary learning and Fisher discrimination sparse representation.By considering the similar sparse representation of features of the same pedestrian in different scenes,the concept of pedestrian re-identification scat- ter function is put forward through adding a regularization term that constrains the sparse representation.The regu- larization term aims at maximizing the between-class scatter of the sparse representation of different pedestrians,and minimizing the within-class scatter of the sparse representation of the same pedestrian.Thus,sparse representation with strong discrimination ability can be obtained via dictionary learning.Experimental results on VIPeR,PRID 450s and CAVIAR4REID datasets indicate that the recognition rate of the proposed method is higher than that of other dictionary learning-based pedestrian re-identification methods.

Key words: pedestrian re-identification, Fisher discrimination, dictionary learning, sparse representation, scatter