Person re-identification using human salience based on multi-feature fusion

Person re-identification plays an important role in matching pedestrians across disjoint camera views. Human salience is distinctive and reliable information in matching, but we will get different results by using different features. In this paper, in order to solve some person reidentification problems, we exploit multi-feature fusion method include RGB, SIFT and Rotation invariant LBP (RI-LBP) to improve the salience feature representation. Due to rotation invariant RI-LBP and SIFT have robust rotation invariant properties, the experiment results are relatively stable. In addition, human salience is also combined with SDALF to improve the performance of person re-identification, and we found a suitable weight between these two methods, which improves the results significantly. Finally, the effectiveness of our approach is validated on the widely used VIPeR dataset, and the experimental results show that our proposed method outperforms most state-of-the-art methods.