Abstract:Person re-identification (Re-ID) technology plays an increasingly crucial role in intelligent surveillance systems. Widespread occlusion significantly impacts the performance of person Re-ID. Occluded person Re-ID refers to a pedestrian matching method that deals with challenges such as pedestrian information loss, noise interference, and perspective misalignment. It has garnered extensive attention from researchers. Over the past few years, several occlusion-solving person Re-ID methods have been proposed, tackling various sub-problems arising from occlusion. However, there is a lack of comprehensive studies that compare, summarize, and evaluate the potential of occluded person Re-ID methods in detail. In this review, we start by providing a detailed overview of the datasets and evaluation scheme used for occluded person Re-ID. Next, we scientifically classify and analyze existing deep learning-based occluded person Re-ID methods from various perspectives, summarizing them concisely. Furthermore, we conduct a systematic comparison among these methods, identify the state-of-the-art approaches, and present an outlook on the future development of occluded person Re-ID.

What problem does this paper attempt to address?

### Problems the paper attempts to solve This paper aims to solve the occlusion problem in person re - identification (Person Re - Identification, Re - ID). Specifically, the occlusion phenomenon is very common in real - world scenarios and can have a serious impact on the performance of person re - identification. Occluded Person Re - ID refers to the method of performing pedestrian matching under challenges such as loss of pedestrian information, noise interference, and misaligned viewpoints. These problems include: 1. **Noise problem**: Due to the interference of multiple mixed information in complex scenes, the feature extraction process will be affected by noise. 2. **Missing problem**: Since only part of the pedestrian area is captured, the pedestrian features are incomplete. 3. **Alignment problem**: Due to changes in posture, viewpoint, and position, features cannot be matched one - by - one, resulting in distraction, shared position misalignment, and other problems. By reviewing existing deep - learning methods, the paper systematically classifies and analyzes these methods and makes a detailed comparison to identify the current state - of - the - art methods and look forward to future development directions. The main contributions of the paper include: 1. **Comprehensive review**: A detailed review of past and current state - of - the - art occluded person re - identification methods. 2. **Introduction of ViT method**: Discussion of methods based on Vision Transformer (ViT) and their hybrid variants, providing new ideas and options for researchers. 3. **Innovative methods**: Creatively incorporating 3D person re - identification and multi - modal person re - identification. These methods utilize additional depth or modal information to better solve the occlusion problem. 4. **Future prospects**: Predicting the future development trend of occluded person re - identification, firmly believing that continuous research and innovation will bring more effective solutions. ### Key point summary - **Problem background**: The occlusion problem is a key challenge in person re - identification, affecting the accuracy and robustness of the system. - **Research method**: The paper classifies and analyzes existing methods from multiple perspectives such as network structure, feature extraction method, and feature hierarchy. - **Main contributions**: A comprehensive review of existing methods, introduction of ViT, 3D, and multi - modal methods, and proposing directions for future research. Through these contributions, the paper provides important references and guidance for research in the field of occluded person re - identification.

Occluded Person Re-Identification with Deep Learning: A Survey and Perspectives

Occluded person re-identification with deep learning: A survey and perspectives

Deep Learning-Based Occluded Person Re-Identification: A Survey

Deep learning-based person re-identification methods: A survey and outlook of recent works

Deep Learning for Person Re-identification: A Survey and Outlook

Deep learning-based methods for person re-identification: A comprehensive review

Person Re-Identification in Special Scenes Based on Deep Learning: A Comprehensive Survey

A Brief Survey for Person Re-Identification Based on Deep Learning.

A State-of-the-art Review on Person Re-Identification with Deep Learning

Occluded Person Re-Identification with Single-scale Global Representations

Adaptive Deep Metric Embeddings for Person Re-Identification under Occlusions

A Semantic Perception and CNN-Transformer Hybrid Network for Occluded Person Re-identification

Person Re-Identification using Deep Learning Networks: A Systematic Review

Semantic-Aware Occlusion-Robust Network for Occluded Person Re-Identification

Person Re-Identification: A Retrospective on Domain Specific Open Challenges and Future Trends

Overview of deep learning based pedestrian attribute recognition and re-identification

A Comprehensive Overview of Person Re-Identification Approaches

Pedestrian 3D Shape Understanding for Person Re-Identification via Multi-View Learning

Occlusion-Sensitive Person Re-Identification Via Attribute-Based Shift Attention

Pose-Guided Feature Learning with Knowledge Distillation for Occluded Person Re-Identification.