Multiple metric learning based on bar-shape descriptor for person re-identification
作者：
Highlights：
• A novel multiple bar-shape structural feature descriptor is proposed to extract robust feature for person re-identification.
• A new multiple metric learning method is proposed from the view of the similarity and relative distance between features. This measure can fully mine the discriminative information and eliminate redundancy of similar features.
• The proposed method is simple but effective for person re-identification.
• Experimental results on popular datasets show that the proposed method is more effective and robust against visual appearance variation.
摘要
•A novel multiple bar-shape structural feature descriptor is proposed to extract robust feature for person re-identification.•A new multiple metric learning method is proposed from the view of the similarity and relative distance between features. This measure can fully mine the discriminative information and eliminate redundancy of similar features.•The proposed method is simple but effective for person re-identification.•Experimental results on popular datasets show that the proposed method is more effective and robust against visual appearance variation.
论文关键词：Person re-identification,Multiple bar-shape descriptor,Multiple metric learning
论文评审过程：Received 4 June 2016, Revised 15 May 2017, Accepted 7 June 2017, Available online 8 June 2017, Version of Record 19 June 2017.
论文官网地址：https://doi.org/10.1016/j.patcog.2017.06.011
