Survey on person re-identification based on deep learning

Abstract

Person re-identification (Re-ID) is a fundamental subject in the field of the computer vision technologies. The traditional methods of person Re-ID have difficulty in solving the problems of person illumination, occlusion and attitude change under complex background. Meanwhile, the introduction of deep learning opens a new way of person Re-ID research and becomes a hot spot in this field. This study reviews the traditional methods of person Re-ID, then the authors focus on the related papers about different person Re-ID frameworks on the basis of deep learning, and discusses their advantages and disadvantages. Finally, they propose the direction of further research, especially the prospect of person Re-ID methods based on deep learning.

Keywords

• computer vision
• learning (artificial intelligence)
• image matching
• image representation
• feature extraction
• deep learning
• computer vision technologies
• person illumination
• occlusion
• attitude change
• person Re-ID research
• person Re-ID methods
• person reidentification
• person Re-ID frameworks
• person matching
• feature-based method
• metric-based method
• person representation